diff --git a/assets/webui/index.html b/assets/webui/index.html
index a78f0f5..964e571 100644
--- a/assets/webui/index.html
+++ b/assets/webui/index.html
@@ -292,7 +292,7 @@
 					<label for="createName">名称</label>
 					<input id="createName" type="text" value="未命名" required autofocus>
 				</div>
-				<div id="createImagesRow" class="form-row" hidden><label for="createImages">图片</label><input id="createImages" type="file" accept=".png,.jpg,.jpeg,.jb2,.jbig2" multiple disabled></div>
+				<div id="createImagesRow" class="form-row" hidden><label for="createImages">图片</label><input id="createImages" type="file" accept=".png,.jpg,.jpeg,.bmp,.tif,.tiff,.jb2,.jbig2" multiple disabled></div>
 				<div id="createSizeRow" class="form-row" hidden><label for="createSize">尺寸</label><select id="createSize" disabled><option value="image">随图</option><option value="custom">自定</option></select></div>
 				<div id="createDimensions" class="form-pair">
 					<div class="form-row">
@@ -466,7 +466,7 @@
 			<form id="insertForm">
 				<div class="form-row">
 					<label for="insertImage">图片</label>
-					<input id="insertImage" type="file" accept=".png,.jpg,.jpeg,image/png,image/jpeg" required autofocus>
+					<input id="insertImage" type="file" accept=".png,.jpg,.jpeg,.bmp,.tif,.tiff,image/png,image/jpeg,image/bmp,image/tiff" required autofocus>
 				</div>
 				<div id="insertFitRow" class="form-row" hidden>
 					<label for="insertFit">缩放</label>
diff --git a/go.mod b/go.mod
index 2d9badb..f390bb3 100644
--- a/go.mod
+++ b/go.mod
@@ -9,7 +9,7 @@ require (
 	github.com/tdewolff/canvas v0.0.0-20260923214215-09804640d00c
 	github.com/xiaoqidun/j2kgo v0.0.0-20261011021706-cedb3037ff6e
 	github.com/xiaoqidun/jbig2 v1.1.4
-	github.com/xiaoqidun/pdfgo v0.0.0-20261011021850-35fe3f96d36f
+	github.com/xiaoqidun/pdfgo v0.0.0-20261011103616-6cafccc7f8f6
 	golang.org/x/crypto v0.58.0
 	golang.org/x/image v0.47.0
 	golang.org/x/text v0.43.0
diff --git a/go.sum b/go.sum
index ddc588e..d1da391 100644
--- a/go.sum
+++ b/go.sum
@@ -62,8 +62,8 @@ github.com/xiaoqidun/j2kgo v0.0.0-20261011021706-cedb3037ff6e h1:OuRcvG5g5AieBN3
 github.com/xiaoqidun/j2kgo v0.0.0-20261011021706-cedb3037ff6e/go.mod h1:DCwAfuDzCDYqUiAlZ7BDwigTjaXw7o9L1FtEma/D1aA=
 github.com/xiaoqidun/jbig2 v1.1.4 h1:LF6J1si4JY+TszTh9+m4KAg9/ba0JjfXoGtQTQC2S8c=
 github.com/xiaoqidun/jbig2 v1.1.4/go.mod h1:9jromqVwMqlptUDEt6Yip8vqYA5I1YjhG7FgMl/Z2dA=
-github.com/xiaoqidun/pdfgo v0.0.0-20261011021850-35fe3f96d36f h1:m3yM1CzrHWm+LW84WWcA2Kr1Ux8K/Znrgd0XVxGEtIw=
-github.com/xiaoqidun/pdfgo v0.0.0-20261011021850-35fe3f96d36f/go.mod h1:hkZn8gHOI94K0fE2ynbXaE8J5T0/jCvoTS/yroKgb+E=
+github.com/xiaoqidun/pdfgo v0.0.0-20261011103616-6cafccc7f8f6 h1:TW7d8qguwR6+WAxXw4ncAXdHTSrAgz6zjj7NmKbipyo=
+github.com/xiaoqidun/pdfgo v0.0.0-20261011103616-6cafccc7f8f6/go.mod h1:hkZn8gHOI94K0fE2ynbXaE8J5T0/jCvoTS/yroKgb+E=
 github.com/xyproto/randomstring v1.0.5 h1:YtlWPoRdgMu3NZtP45drfy1GKoojuR7hmRcnhZqKjWU=
 github.com/xyproto/randomstring v1.0.5/go.mod h1:rgmS5DeNXLivK7YprL0pY+lTuhNQW3iGxZ18UQApw/E=
 github.com/yuin/goldmark v1.8.5 h1:r6N5afV5qj/5S4UTch8agZHJ8UxNCMwX7WjkkJam2NA=
diff --git a/internal/webui/wasm.go b/internal/webui/wasm.go
index 5abe833..c4278a9 100644
--- a/internal/webui/wasm.go
+++ b/internal/webui/wasm.go
@@ -25,7 +25,6 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"image"
 	"image/color"
 	"io"
 	"io/fs"
@@ -1570,11 +1569,11 @@ func writeAnnotation(args []js.Value) (any, error) {
 			if err != nil {
 				return err
 			}
-			config, _, err := image.DecodeConfig(bytes.NewReader(data))
+			size, err := currentEditor.ImageSize(resource)
 			if err != nil {
 				return err
 			}
-			options.Height = options.Width * float64(config.Height) / float64(config.Width)
+			options.Height = options.Width * float64(size.Y) / float64(size.X)
 		}
 		if options.Kind == "replace" {
 			style := ofdgo.TextStyle{Font: font}
@@ -3297,15 +3296,16 @@ func importPages(args []js.Value) (any, error) {
 	}
 	at := args[1].Int()
 	options := ofdgo.PageImportOptions{Outlines: len(args) > 2 && args[2].Bool()}
+	ctx := context.Background()
 	progressIndex := 3
 	if len(args) > 3 && args[3].Type() == js.TypeBoolean {
 		options.AllowDecrypted = args[3].Bool()
 		progressIndex = 4
 	}
 	if len(args) > progressIndex {
-		options.OnProgress = editorOperationProgress(args[progressIndex])
+		ctx, options.OnProgress = newEditorOperationContext(args[progressIndex])
 	}
-	if _, err := currentEditor.ImportPagesWithOptions(pendingImport, indexes, at, options); err != nil {
+	if _, err := currentEditor.ImportPagesContext(ctx, pendingImport, indexes, at, options); err != nil {
 		return nil, err
 	}
 	clearImport()
@@ -4103,12 +4103,12 @@ func insertImage(args []js.Value) (any, error) {
 	if err != nil {
 		return nil, err
 	}
-	config, _, err := image.DecodeConfig(bytes.NewReader(data))
+	size, err := currentEditor.ImageSize(resourceID)
 	if err != nil {
 		return nil, err
 	}
 	width := args[4].Float()
-	box := ofdgo.Box{X: args[2].Float(), Y: args[3].Float(), W: width, H: width * float64(config.Height) / float64(config.Width)}
+	box := ofdgo.Box{X: args[2].Float(), Y: args[3].Float(), W: width, H: width * float64(size.Y) / float64(size.X)}
 	content, err := currentSession.pageContent(args[0].Int())
 	if err != nil {
 		return nil, err
@@ -4593,11 +4593,11 @@ func compressionFromJS(value js.Value, options *ofdgo.CompressionOptions) error
 	return options.Validate()
 }
 
-// outputContext 定期让出浏览器线程，将取消请求传入库层压缩循环
+// outputContext 定期让出浏览器线程，将取消请求传入库层长操作
 type outputContext struct {
 	context.Context
 	cancel     context.CancelFunc
-	checkpoint js.Value
+	checkpoint func() error
 	checked    time.Time
 }
 
@@ -4606,7 +4606,21 @@ type outputContext struct {
 // 返回: context.Context 导出上下文
 func newOutputContext(checkpoint js.Value) context.Context {
 	ctx, cancel := context.WithCancel(context.Background())
-	return &outputContext{Context: ctx, cancel: cancel, checkpoint: checkpoint}
+	return &outputContext{Context: ctx, cancel: cancel, checkpoint: func() error { return awaitExport(checkpoint) }}
+}
+
+// newEditorOperationContext 在底层取消检查时保持最近的进度，避免版本处理期间界面无响应
+// 入参: callback 浏览器进度及取消回调
+// 返回: context.Context 操作上下文, func(string, int, int) error 阶段进度回调
+func newEditorOperationContext(callback js.Value) (context.Context, func(string, int, int) error) {
+	ctx, cancel := context.WithCancel(context.Background())
+	stage, completed, total := "ids", 0, 0
+	progress := editorOperationProgress(callback)
+	operation := &outputContext{Context: ctx, cancel: cancel, checkpoint: func() error { return progress(stage, completed, total) }}
+	return operation, func(next string, done, count int) error {
+		stage, completed, total = next, done, count
+		return progress(stage, completed, total)
+	}
 }
 
 // Err 在取消检查间隙处理浏览器消息，不为每个像素触发跨语言回调
@@ -4617,7 +4631,7 @@ func (c *outputContext) Err() error {
 	}
 	if time.Since(c.checked) >= 50*time.Millisecond {
 		c.checked = time.Now()
-		if err := awaitExport(c.checkpoint); err != nil {
+		if err := c.checkpoint(); err != nil {
 			c.cancel()
 		}
 	}
diff --git a/ofdgo_compression.go b/ofdgo_compression.go
index fe7296d..0c80b2f 100644
--- a/ofdgo_compression.go
+++ b/ofdgo_compression.go
@@ -101,14 +101,19 @@ func (e *Editor) WriteToWithOptions(ctx context.Context, writer io.Writer, optio
 	return snapshot.WriteTo(convertWriter{context: ctx, writer: writer, count: &count})
 }
 
-// WritePagesToWithOptions 按指定顺序另存页面并应用输出压缩
+// WritePagesToWithOptions 按指定顺序另存页面，仅为所选页面准备资源和压缩计划
 // 入参: ctx 取消上下文, writer 输出流, indexes 零基页码, options 输出配置
 // 返回: int64 已写入字节数, error 错误信息
 func (e *Editor) WritePagesToWithOptions(ctx context.Context, writer io.Writer, indexes []int, options WriteOptions) (int64, error) {
-	snapshot, err := e.outputSnapshot(ctx, options)
-	if err != nil {
+	if err := ctx.Err(); err != nil {
 		return 0, err
 	}
+	if err := options.Compression.Validate(); err != nil {
+		return 0, err
+	}
+	snapshot := *e
+	snapshot.OnWriteProgress = editorProgress(e.OnWriteProgress).withContext(ctx)
+	snapshot.output = &outputOptimization{ctx: ctx, options: options.Compression}
 	var count int64
 	return snapshot.WritePagesTo(convertWriter{context: ctx, writer: writer, count: &count}, indexes)
 }
@@ -130,15 +135,10 @@ func (e *Editor) outputSnapshot(ctx context.Context, options WriteOptions) (*Edi
 		}
 		return editorProgress(e.OnWriteProgress).report(stage, completed, total)
 	}
-	snapshot.output = &outputOptimization{ctx: ctx, options: options.Compression, images: map[string]bool{}}
+	snapshot.output = &outputOptimization{ctx: ctx, options: options.Compression}
 	if options.Compression.Mode == CompressionUnchanged {
 		return &snapshot, nil
 	}
-	for _, resource := range e.resources {
-		if resource.image != nil {
-			snapshot.output.images[cleanPackagePath(resource.name)] = true
-		}
-	}
 	if e.source != nil {
 		protected, err := e.source.reader.hasOutputSignatures()
 		if err != nil {
@@ -146,19 +146,11 @@ func (e *Editor) outputSnapshot(ctx context.Context, options WriteOptions) (*Edi
 		}
 		snapshot.output.protected = protected
 		if protected {
-			snapshot.output.images = nil
 			return &snapshot, nil
 		}
-		images, err := e.source.reader.Images(ctx)
-		if err != nil {
-			return nil, err
-		}
-		for _, img := range images {
-			snapshot.output.images[cleanPackagePath(img.Location)] = true
-		}
 	}
 	if options.Compression.Mode == CompressionLossy {
-		reader, err := e.reader(editorProgress(snapshot.OnWriteProgress))
+		reader, err := e.readerContext(ctx, editorProgress(snapshot.OnWriteProgress))
 		if err != nil {
 			return nil, err
 		}
@@ -174,7 +166,7 @@ func (e *Editor) outputSnapshot(ctx context.Context, options WriteOptions) (*Edi
 	return &snapshot, nil
 }
 
-// compressionImageSafety 保护蒙版及其原图和替换图像，识别不能可靠降采样的图案及隐藏内容
+// compressionImageSafety 保护蒙版及其原图和替换图像，不完整的文档结构不参与降采样
 // 入参: ctx 取消上下文, images 图片资源
 // 返回: map[string]bool 只能无损处理的资源路径, bool 是否可分析尺寸, error 读取或取消错误
 func (r *Reader) compressionImageSafety(ctx context.Context, images []ImageInfo) (map[string]bool, bool, error) {
@@ -205,7 +197,6 @@ func (r *Reader) compressionImageSafety(ctx context.Context, images []ImageInfo)
 		}
 	}
 	masks := make(map[string]bool)
-	geometry := true
 	for _, name := range slices.Sorted(maps.Keys(names)) {
 		if !safe {
 			break
@@ -248,17 +239,11 @@ func (r *Reader) compressionImageSafety(ctx context.Context, images []ImageInfo)
 						break
 					}
 					stack = append(stack, element.Name)
-					if element.Name.Local == "Pattern" {
-						geometry = false
-					}
 					resource, substitution, masked := "", "", false
 					for _, attr := range element.Attr {
 						if attr.Name.Space != "" {
 							continue
 						}
-						if attr.Name.Local == "Visible" && (attr.Value == "false" || attr.Value == "0") {
-							geometry = false
-						}
 						if attr.Name.Local == "ResourceID" {
 							resource = editorResourceID(attr.Value)
 						}
@@ -278,6 +263,8 @@ func (r *Reader) compressionImageSafety(ctx context.Context, images []ImageInfo)
 					}
 				case xml.EndElement:
 					stack = stack[:len(stack)-1]
+				case xml.Directive:
+					safe = false
 				}
 				if !safe {
 					break
@@ -295,7 +282,7 @@ func (r *Reader) compressionImageSafety(ctx context.Context, images []ImageInfo)
 			paths[cleanPackagePath(img.Location)] = true
 		}
 	}
-	return paths, safe && geometry, ctx.Err()
+	return paths, safe, ctx.Err()
 }
 
 // hasOutputSignatures 检查包中各文档，避免改写其他文档签名保护的共享资源
diff --git a/ofdgo_compression_images.go b/ofdgo_compression_images.go
index 873585a..cb60d0e 100644
--- a/ofdgo_compression_images.go
+++ b/ofdgo_compression_images.go
@@ -23,15 +23,22 @@ import (
 
 // compressionImages 记录图片各处使用的最大像素需求
 type compressionImages struct {
-	ctx   context.Context
-	dpi   int
-	sizes map[string]image.Point
+	ctx           context.Context
+	dpi           int
+	sizes         map[string]image.Point
+	renderer      *Renderer
+	patternPage   *Matrix
+	patterns      map[*Pattern]bool
+	patternVisits map[compressionPatternKey]bool
 }
 
-// compressionImagePlan 分文档分析标识，按共享路径合并最大尺寸和保护条件
+// compressionImagePlan 为有损输出分文档分析图片，按共享路径合并最大尺寸和保护条件
 // 入参: ctx 取消上下文, options 压缩配置
 // 返回: map[string]bool 可有损处理的路径, map[string]image.Point 像素需求, error 读取或取消错误
 func (r *Reader) compressionImagePlan(ctx context.Context, options CompressionOptions) (map[string]bool, map[string]image.Point, error) {
+	if options.Mode != CompressionLossy {
+		return nil, nil, ctx.Err()
+	}
 	images := make(map[string]bool)
 	sizes := make(map[string]image.Point)
 	blocked := make(map[string]bool)
@@ -54,14 +61,10 @@ func (r *Reader) compressionImagePlan(ctx context.Context, options CompressionOp
 			}
 			return nil, nil, err
 		}
-		var masks map[string]bool
 		var demands map[string]image.Point
-		if options.Mode == CompressionLossy {
-			var geometry bool
-			masks, geometry, err = document.compressionImageSafety(ctx, catalog)
-			if err == nil && geometry {
-				demands, err = document.compressionImageSizes(ctx, catalog, options.ImageDPI())
-			}
+		masks, geometry, err := document.compressionImageSafety(ctx, catalog)
+		if err == nil && geometry {
+			demands, err = document.compressionImageSizes(ctx, catalog, options.ImageDPI())
 		}
 		if document != r {
 			document.Close()
@@ -102,7 +105,7 @@ func compressionImageSize(width, height float64, dpi int) image.Point {
 	return image.Pt(int(w), int(h))
 }
 
-// compressionImageSizes 按正文、模板及注解的实际变换汇总共享图片尺寸
+// compressionImageSizes 按正文、模板、注解及底纹的实际变换汇总图片尺寸，包括隐藏内容
 // 入参: ctx 取消上下文, images 图片资源, dpi 分辨率上限
 // 返回: map[string]image.Point 资源像素需求, error 取消错误
 func (r *Reader) compressionImageSizes(ctx context.Context, images []ImageInfo, dpi int) (map[string]image.Point, error) {
@@ -113,9 +116,9 @@ func (r *Reader) compressionImageSizes(ctx context.Context, images []ImageInfo,
 	if err != nil {
 		return nil, ctx.Err()
 	}
-	renderer := NewRenderer(r)
+	renderer := NewRenderer(r, WithRenderBackends(RenderBackends{}))
 	defer renderer.ClearCache()
-	visitor := &compressionImages{ctx: ctx, dpi: dpi, sizes: make(map[string]image.Point)}
+	visitor := &compressionImages{ctx: ctx, dpi: dpi, sizes: make(map[string]image.Point), renderer: renderer}
 	for index := 0; index < count; index++ {
 		if err := ctx.Err(); err != nil {
 			return nil, err
@@ -157,14 +160,21 @@ func (r *Reader) compressionImageSizes(ctx context.Context, images []ImageInfo,
 	return result, ctx.Err()
 }
 
-// DrawObject 汇总原图和替换图像的完整显示尺寸，不使用裁剪后的局部面积
+// DrawObject 汇总图片、替换图及文字和路径底纹的尺寸，不使用裁剪后的局部面积
 // 入参: object 图形对象, state 继承状态
 // 返回: error 取消或几何错误
 func (v *compressionImages) DrawObject(object *GraphicObject, state RenderState) error {
 	if err := v.ctx.Err(); err != nil {
 		return err
 	}
-	if object.Type != "ImageObject" {
+	switch object.Type {
+	case "PathObject":
+		return v.objectPatterns(object.PathObject, state)
+	case "TextObject":
+		text := object.TextObject
+		return v.objectPatterns(PathObject{Boundary: text.Boundary, CTM: text.CTM, DrawParam: text.DrawParam, FillColor: text.FillColor, StrokeColor: text.StrokeColor}, state)
+	case "ImageObject":
+	default:
 		return nil
 	}
 	img := object.ImageObject
@@ -182,6 +192,35 @@ func (v *compressionImages) DrawObject(object *GraphicObject, state RenderState)
 	}
 	matrix, _ := renderObjectMatrix(img.Boundary, local, state)
 	size := compressionImageSize(math.Hypot(matrix.a, matrix.b), math.Hypot(matrix.c, matrix.d), v.dpi)
+	return v.recordSize(ids, size)
+}
+
+// VisitComposite 按矢量资源的完整变换统计缩略图和替换图，不以裁剪区域缩减尺寸
+// 入参: object 矢量资源或复合图元, state 已包含当前图元变换的状态
+// 返回: error 取消、资源标识或几何错误
+func (v *compressionImages) VisitComposite(object *CompositeGraphicUnit, state RenderState) error {
+	if err := v.ctx.Err(); err != nil {
+		return err
+	}
+	if object.Thumbnail == "" && object.Substitution == "" {
+		return nil
+	}
+	ids := [2]string{editorResourceID(object.Thumbnail), editorResourceID(object.Substitution)}
+	if object.Thumbnail != "" && ids[0] == "" || object.Substitution != "" && ids[1] == "" {
+		return fmt.Errorf("invalid composite image resource")
+	}
+	matrix := IdentityMatrix
+	if state.Parent != nil {
+		matrix = *state.Parent
+	}
+	size := compressionImageSize(object.Width*math.Hypot(matrix.a, matrix.b), object.Height*math.Hypot(matrix.c, matrix.d), v.dpi)
+	return v.recordSize(ids, size)
+}
+
+// recordSize 合并同一图像资源在不同位置的最大像素需求
+// 入参: ids 图像标识，空值不统计, size 当前像素需求
+// 返回: error 尺寸无效时返回错误
+func (v *compressionImages) recordSize(ids [2]string, size image.Point) error {
 	if size.X == 0 || size.Y == 0 {
 		return fmt.Errorf("invalid image placement")
 	}
@@ -199,3 +238,11 @@ func (v *compressionImages) DrawObject(object *GraphicObject, state RenderState)
 // 入参: stamp 印章
 // 返回: error 始终为空
 func (v *compressionImages) DrawStamp(stamp Stamp) error { return nil }
+
+// IncludeHidden 将隐藏内容纳入尺寸分析，保留图片在所有使用位置的精度
+// 返回: bool 始终为真
+func (v *compressionImages) IncludeHidden() bool { return true }
+
+// SkipClips 按完整图片尺寸分析，不加载裁剪几何或文字轮廓
+// 返回: bool 始终为真
+func (v *compressionImages) SkipClips() bool { return true }
diff --git a/ofdgo_compression_parallel.go b/ofdgo_compression_parallel.go
index 329ec37..e898d55 100644
--- a/ofdgo_compression_parallel.go
+++ b/ofdgo_compression_parallel.go
@@ -15,13 +15,10 @@
 package ofdgo
 
 import (
-	"bytes"
 	"crypto/sha256"
 	"image"
 	"runtime"
 	"sync"
-
-	"github.com/xiaoqidun/pdfgo"
 )
 
 // outputImageBatchLimit 限制并行批次持有的原始编码，超限图片单独处理
@@ -37,8 +34,8 @@ type outputImageJob struct {
 
 // outputImageSource 保存已读取资源的摘要和格式，不保留原始编码
 type outputImageSource struct {
-	key  outputImageKey
-	jpeg bool
+	key    outputImageKey
+	format string
 }
 
 // outputImageOptions 统一资源压缩及像素需求，受保护图片只进行无损编码
@@ -92,7 +89,7 @@ func (e *Editor) prepareOutputImages(keys []string, load func(string) ([]byte, e
 		options, demand := e.outputImageOptions(key)
 		cacheKey := outputImageKey{hash: sha256.Sum256(data), options: options, size: demand}
 		if sources != nil {
-			sources[key] = outputImageSource{key: cacheKey, jpeg: bytes.HasPrefix(data, []byte{255, 216})}
+			sources[key] = outputImageSource{key: cacheKey, format: outputImageFormat(data)}
 		}
 		if _, cached := results[cacheKey]; cached {
 			continue
@@ -126,7 +123,7 @@ func (e *Editor) prepareOutputImages(keys []string, load func(string) ([]byte, e
 func (e *Editor) runOutputImages(jobs []outputImageJob, results map[outputImageKey][]byte) error {
 	ctx := e.output.ctx
 	optimize := func(job *outputImageJob) {
-		job.candidate, job.err = pdfgo.OptimizeImageResource(ctx, job.data, job.key.options, job.key.size)
+		job.candidate, job.err = optimizeImageResource(ctx, job.data, job.key.options, job.key.size)
 	}
 	if len(jobs) == 1 {
 		optimize(&jobs[0])
diff --git a/ofdgo_compression_pattern_images.go b/ofdgo_compression_pattern_images.go
new file mode 100644
index 0000000..eb8f667
--- /dev/null
+++ b/ofdgo_compression_pattern_images.go
@@ -0,0 +1,142 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"fmt"
+	"math"
+)
+
+// compressionPatternImageLimit 限制底纹尺寸分析的变换组合数，超限时保留原图尺寸
+const compressionPatternImageLimit = 65536
+
+// compressionPatternKey 复用相同底纹的尺寸分析，平移不影响像素需求
+type compressionPatternKey struct {
+	pattern *Pattern
+	linear  [4]float64
+}
+
+// objectPatterns 统计文字和路径的填充及描边底纹，保留绘制参数继承关系
+// 入参: object 对象样式及坐标, state 继承状态
+// 返回: error 底纹分析错误
+func (v *compressionImages) objectPatterns(object PathObject, state RenderState) error {
+	fill, stroke := object.FillColor, (*FillColor)(object.StrokeColor)
+	if defaults := v.renderer.drawParamDefaults(object.DrawParam, state.Defaults); defaults != nil {
+		if fill == nil {
+			fill = defaults.FillColor
+		}
+		if stroke == nil {
+			stroke = (*FillColor)(defaults.StrokeColor)
+		}
+	}
+	if (fill == nil || fill.Pattern == nil) && (stroke == nil || stroke.Pattern == nil) {
+		return nil
+	}
+	matrix, _ := renderObjectMatrix(object.Boundary, NewMatrix(object.CTM), state)
+	for _, color := range [2]*FillColor{fill, stroke} {
+		if color != nil && color.Pattern != nil {
+			if err := v.patternImages(color, matrix); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
+// patternImages 按独立单元坐标汇总底纹图片及缩略图，不展开重复铺排的单元
+// 入参: color 底纹颜色, object 对象到页面矩阵
+// 返回: error 取消、循环引用或变换错误
+func (v *compressionImages) patternImages(color *FillColor, object Matrix) error {
+	if err := v.ctx.Err(); err != nil {
+		return err
+	}
+	pattern := color.Pattern
+	if v.patterns[pattern] {
+		return fmt.Errorf("cyclic pattern image reference")
+	}
+	if !finite(pattern.Width) || !finite(pattern.Height) || pattern.Width <= 0 || pattern.Height <= 0 {
+		return fmt.Errorf("invalid pattern cell size")
+	}
+	matrix := NewMatrix(pattern.CTM)
+	switch pattern.RelativeTo {
+	case "", "Object":
+		matrix = object.Multiply(matrix)
+	case "Page":
+		if v.patternPage != nil {
+			matrix = v.patternPage.Multiply(matrix)
+		}
+	default:
+		return fmt.Errorf("invalid pattern coordinate system")
+	}
+	if !finite(matrix.a) || !finite(matrix.b) || !finite(matrix.c) || !finite(matrix.d) {
+		return fmt.Errorf("invalid pattern transform")
+	}
+	matrix.e, matrix.f = 0, 0
+	key := compressionPatternKey{pattern: pattern, linear: [4]float64{matrix.a, matrix.b, matrix.c, matrix.d}}
+	if v.patternVisits[key] {
+		return nil
+	}
+	if len(v.patternVisits) >= compressionPatternImageLimit {
+		return fmt.Errorf("pattern image analysis limit exceeded")
+	}
+	if v.patterns == nil {
+		v.patterns = make(map[*Pattern]bool)
+		v.patternVisits = make(map[compressionPatternKey]bool)
+	}
+	v.patternVisits[key] = true
+	v.patterns[pattern] = true
+	defer delete(v.patterns, pattern)
+	if pattern.CellContent.Thumbnail != "" {
+		id := editorResourceID(pattern.CellContent.Thumbnail)
+		if id == "" {
+			return fmt.Errorf("invalid pattern thumbnail resource")
+		}
+		size := compressionImageSize(pattern.Width*math.Hypot(matrix.a, matrix.b), pattern.Height*math.Hypot(matrix.c, matrix.d), v.dpi)
+		if err := v.recordSize([2]string{id}, size); err != nil {
+			return err
+		}
+	}
+	columns, rows := 1, 1
+	switch pattern.ReflectMethod {
+	case "", "Normal":
+	case "Column":
+		columns = 2
+	case "Row":
+		rows = 2
+	case "RowAndColumn":
+		columns, rows = 2, 2
+	default:
+		return fmt.Errorf("invalid pattern reflection")
+	}
+	parent := v.patternPage
+	defer func() { v.patternPage = parent }()
+	solid := FillColor{Value: color.Value, Index: color.Index, ColorSpace: color.ColorSpace}
+	defaults := &DrawParam{FillColor: &solid, StrokeColor: (*StrokeColor)(&solid)}
+	for column := range columns {
+		for row := range rows {
+			cell := matrix.Multiply(Matrix{a: float64(1 - 2*column), d: float64(1 - 2*row)})
+			v.patternPage = &cell
+			for member := range pattern.CellContent.objects() {
+				if err := v.ctx.Err(); err != nil {
+					return err
+				}
+				if err := v.renderer.WalkObject(&member, RenderState{Defaults: defaults, Parent: &cell, BoundaryInCTM: true}, v); err != nil {
+					return err
+				}
+			}
+		}
+	}
+	return nil
+}
diff --git a/ofdgo_compression_profiles.go b/ofdgo_compression_profiles.go
new file mode 100644
index 0000000..1e6879b
--- /dev/null
+++ b/ofdgo_compression_profiles.go
@@ -0,0 +1,135 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"crypto/sha256"
+	"fmt"
+	"io"
+	"maps"
+	"slices"
+)
+
+// editorProfileReference 记录颜色配置文件所在的资源索引及颜色空间标识
+type editorProfileReference struct {
+	index string
+	id    string
+}
+
+// compactResourceProfiles 合并相同颜色配置文件，保留颜色空间声明和独立文件引用
+// 入参: ctx 取消上下文, reader 原包，可为空, parts 输出改动, profiles 配置文件引用, protected 不可替换的文件, removed 待移除文件
+// 返回: error 读取、改写或取消错误
+func compactResourceProfiles(ctx context.Context, reader *Reader, parts map[string][]byte, profiles map[string][]editorProfileReference, protected, removed map[string]bool) error {
+	if len(profiles) < 2 {
+		return ctx.Err()
+	}
+	seen := make(map[[32]byte]string)
+	aliases := make(map[string]string)
+	indexes := make(map[string]map[string]string)
+	buffer := make([]byte, 32*1024)
+	for _, name := range slices.Sorted(maps.Keys(profiles)) {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+		if protected[name] {
+			continue
+		}
+		var input io.ReadCloser
+		if data, ok := parts[name]; ok {
+			input = io.NopCloser(bytes.NewReader(data))
+		} else {
+			if reader == nil {
+				return fmt.Errorf("file not found: %s", name)
+			}
+			var err error
+			input, err = reader.openFile(name)
+			if err != nil {
+				return err
+			}
+		}
+		hash := sha256.New()
+		_, err := io.CopyBuffer(hash, imageInput{ReadCloser: input, context: ctx}, buffer)
+		closeErr := input.Close()
+		if err != nil {
+			return err
+		}
+		if closeErr != nil {
+			return closeErr
+		}
+		key := [32]byte(hash.Sum(nil))
+		first, found := seen[key]
+		if !found {
+			seen[key] = name
+			continue
+		}
+		aliases[name] = first
+		for _, reference := range profiles[name] {
+			if indexes[reference.index] == nil {
+				indexes[reference.index] = make(map[string]string)
+			}
+			indexes[reference.index][reference.id] = first
+		}
+	}
+	updates := make(map[string][]byte, len(indexes))
+	for _, name := range slices.Sorted(maps.Keys(indexes)) {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+		data, ok := parts[name]
+		if !ok {
+			if reader == nil {
+				return fmt.Errorf("file not found: %s", name)
+			}
+			var err error
+			data, err = reader.readFile(name)
+			if err != nil {
+				return err
+			}
+		}
+		root, err := parseEditorXMLContext(ctx, data)
+		if err != nil {
+			return err
+		}
+		var patches []editorXMLPatch
+		for _, group := range root.children {
+			if !group.matchesOFD("ColorSpaces") {
+				continue
+			}
+			for _, node := range group.children {
+				location := indexes[name][editorResourceID(node.attr("ID"))]
+				if !node.matchesOFD("ColorSpace") || location == "" {
+					continue
+				}
+				fragment, err := editorXMLAttribute(data, node, "Profile", "/"+location)
+				if err != nil {
+					return err
+				}
+				patches = append(patches, editorXMLPatch{start: node.start, end: node.end, data: fragment})
+			}
+		}
+		updates[name] = editorPatchXML(data, patches)
+	}
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+	maps.Copy(parts, updates)
+	for name := range aliases {
+		removed[name] = true
+		delete(parts, name)
+	}
+	return nil
+}
diff --git a/ofdgo_compression_raster.go b/ofdgo_compression_raster.go
new file mode 100644
index 0000000..a05e2e9
--- /dev/null
+++ b/ofdgo_compression_raster.go
@@ -0,0 +1,183 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"encoding/binary"
+	"hash/crc32"
+	"image"
+	"io"
+	"math"
+
+	"github.com/xiaoqidun/pdfgo"
+)
+
+// rasterDensity 保存横纵像素密度，unit为0表示比例、1表示每米像素数
+type rasterDensity struct {
+	x, y float64
+	unit byte
+}
+
+// outputImageFormat 识别可由资源压缩生成的图片格式
+// 入参: data 图片编码
+// 返回: string 标准资源格式，未识别时为空
+func outputImageFormat(data []byte) string {
+	if isJPEGData(data) {
+		return "JPEG"
+	}
+	if isPNGData(data) {
+		return "PNG"
+	}
+	return ""
+}
+
+// optimizeImageResource 将可完整保留的BMP及TIFF转为PNG，再复用图像压缩和降采样策略
+// 原始模式及无法保留的色彩配置不转换，仅采用小于原始编码的结果
+// 入参: ctx 取消上下文, data 图片编码, options 压缩选项, size 像素需求
+// 返回: []byte 图片编码，可能引用输入, error 解码、编码或取消错误
+func optimizeImageResource(ctx context.Context, data []byte, options CompressionOptions, size image.Point) ([]byte, error) {
+	if err := options.Validate(); err != nil {
+		return nil, err
+	}
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	if options.Mode != CompressionUnchanged && isTIFFData(data) {
+		return optimizeTIFFResource(ctx, data, options, size)
+	}
+	if options.Mode == CompressionUnchanged || !bytes.HasPrefix(data, []byte("BM")) {
+		return pdfgo.OptimizeImageResource(ctx, data, options, size)
+	}
+	return optimizeBMPResource(ctx, data, options, size)
+}
+
+// optimizeBMPResource 压缩可保留色彩与密度的位图，未知元数据及高精度位域保持原样
+// 入参: ctx 取消上下文, data BMP编码, options 压缩选项, size 像素需求
+// 返回: []byte 图片编码, error 解码、编码或取消错误
+func optimizeBMPResource(ctx context.Context, data []byte, options CompressionOptions, size image.Point) ([]byte, error) {
+	offset, width, height, depth, compression, err := parseBMPHeader(data)
+	if err != nil {
+		return data, nil
+	}
+	if height < 0 {
+		height = -height
+	}
+	if int64(width)*int64(height) > outputOptimizationBufferLimit/4 || compression != 0 && compression != 3 {
+		return data, nil
+	}
+	header := binary.LittleEndian.Uint32(data[14:18])
+	if header != 40 && header != 108 && header != 124 {
+		return data, nil
+	}
+	if header >= 108 && binary.LittleEndian.Uint32(data[70:74]) != 0x73524742 {
+		return data, nil
+	}
+	if header == 124 && (binary.LittleEndian.Uint32(data[122:126]) != 0 || binary.LittleEndian.Uint32(data[126:130]) != 0 || binary.LittleEndian.Uint32(data[130:134]) != 0) {
+		return data, nil
+	}
+	stride := (int64(width)*int64(depth) + 31) / 32 * 4
+	if int64(offset)+stride*int64(height) != int64(len(data)) {
+		return data, nil
+	}
+	xdpi, ydpi := int32(binary.LittleEndian.Uint32(data[38:42])), int32(binary.LittleEndian.Uint32(data[42:46]))
+	if xdpi < 0 || ydpi < 0 {
+		return data, nil
+	}
+	if depth == 16 || depth == 32 {
+		masks, err := bmpColorMasks(data, offset, depth, compression)
+		if err != nil {
+			return data, nil
+		}
+		if options.Mode == CompressionLossless {
+			for _, mask := range masks {
+				if mask.max > 255 {
+					return data, nil
+				}
+			}
+		}
+	}
+	var pixels image.Image
+	if depth == 16 || compression == 3 {
+		pixels, err = decodeBMPImageContext(ctx, data)
+	} else {
+		pixels, _, err = image.Decode(imageInput{ReadCloser: io.NopCloser(bytes.NewReader(data)), context: ctx})
+	}
+	if err != nil {
+		return nil, err
+	}
+	return optimizeRasterResource(ctx, data, pixels, options, size, rasterDensity{x: float64(xdpi), y: float64(ydpi), unit: 1})
+}
+
+// optimizeRasterResource 编码已解码的资源图片，同步降采样后的像素密度，仅采用更小结果
+// 入参: ctx 取消上下文, data 原编码, pixels 像素, options 压缩选项, size 像素需求, density 像素密度
+// 返回: []byte 图片编码, error 编码或取消错误
+func optimizeRasterResource(ctx context.Context, data []byte, pixels image.Image, options CompressionOptions, size image.Point, density rasterDensity) ([]byte, error) {
+	width, height := pixels.Bounds().Dx(), pixels.Bounds().Dy()
+	var encoded bytes.Buffer
+	limit := &outputCompressionBuffer{buffer: &encoded, limit: outputOptimizationBufferLimit}
+	if err := pdfgo.EncodePNG(ctx, limit, pixels); err != nil {
+		if limit.exceeded {
+			return data, ctx.Err()
+		}
+		return nil, err
+	}
+	var candidate []byte
+	var err error
+	if density.x == 0 && density.y == 0 {
+		candidate, err = pdfgo.OptimizeImageResource(ctx, encoded.Bytes(), options, size)
+	} else {
+		candidate, err = pdfgo.OptimizeImageSize(ctx, encoded.Bytes(), options, size)
+		if err == nil {
+			config, _, configErr := image.DecodeConfig(bytes.NewReader(candidate))
+			if configErr != nil {
+				return nil, configErr
+			}
+			x := density.x * float64(config.Width) / float64(width)
+			y := density.y * float64(config.Height) / float64(height)
+			if density.unit == 0 {
+				factor := float64(math.MaxUint32) / max(x, y)
+				x, y = x*factor, y*factor
+			}
+			x, y = math.Round(x), math.Round(y)
+			if x > math.MaxUint32 || y > math.MaxUint32 || density.x > 0 && x < 1 || density.y > 0 && y < 1 {
+				return data, ctx.Err()
+			}
+			chunk := make([]byte, 21)
+			binary.BigEndian.PutUint32(chunk, 9)
+			copy(chunk[4:], "pHYs")
+			binary.BigEndian.PutUint32(chunk[8:], uint32(x))
+			binary.BigEndian.PutUint32(chunk[12:], uint32(y))
+			chunk[16] = density.unit
+			binary.BigEndian.PutUint32(chunk[17:], crc32.ChecksumIEEE(chunk[4:17]))
+			var output bytes.Buffer
+			writer := imageMetadataWriter{writer: &output, skip: 33, data: chunk}
+			if _, err = writer.Write(candidate); err == nil {
+				candidate = output.Bytes()
+			}
+		}
+	}
+	if err != nil {
+		return nil, err
+	}
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	if len(candidate) < len(data) {
+		return candidate, nil
+	}
+	return data, nil
+}
diff --git a/ofdgo_compression_resources.go b/ofdgo_compression_resources.go
index fed45bf..6f8aa31 100644
--- a/ofdgo_compression_resources.go
+++ b/ofdgo_compression_resources.go
@@ -24,8 +24,6 @@ import (
 	"path"
 	"slices"
 	"strings"
-
-	"github.com/xiaoqidun/pdfgo"
 )
 
 // outputImageKey 区分相同图像的压缩配置和像素需求，不跨写出保留状态
@@ -39,6 +37,13 @@ type outputImageKey struct {
 // 入参: parts 输出改动, reader 原包，可为空, removed 已移除条目
 // 返回: error 读取或取消错误
 func (e *Editor) compressResourceParts(parts map[string][]byte, reader *Reader, removed map[string]bool) error {
+	if reader == nil && e.output != nil && e.output.options.Mode != CompressionUnchanged && !e.output.protected {
+		unused, err := e.compactResourceReferences(parts, nil, editorProgress(e.OnWriteProgress), nil, nil)
+		if err != nil {
+			return err
+		}
+		maps.Copy(removed, unused)
+	}
 	return e.compressResourceReferences(parts, reader, removed, nil)
 }
 
@@ -225,7 +230,7 @@ func (e *Editor) compressResourceReferences(parts map[string][]byte, reader *Rea
 				candidate, cached := imageResults[cacheKey]
 				var err error
 				if !cached {
-					candidate, err = pdfgo.OptimizeImageResource(e.output.ctx, data, options, demand)
+					candidate, err = optimizeImageResource(e.output.ctx, data, options, demand)
 					if err == nil {
 						imageResults[cacheKey] = nil
 						if len(candidate) < len(data) {
@@ -245,12 +250,16 @@ func (e *Editor) compressResourceReferences(parts map[string][]byte, reader *Rea
 					source.key.hash = sha256.Sum256(data)
 				}
 				if data != nil {
-					source.jpeg = bytes.HasPrefix(data, []byte{255, 216})
+					source.format = outputImageFormat(data)
 				}
-				if source.jpeg {
-					formats[key] = "JPEG"
+				if source.format != "" {
+					formats[key] = source.format
+					suffix := ".png"
+					if source.format == "JPEG" {
+						suffix = ".jpg"
+					}
 					ext := path.Ext(actual[key])
-					if !strings.EqualFold(ext, ".jpg") && !strings.EqualFold(ext, ".jpeg") {
+					if !strings.EqualFold(ext, suffix) && !(source.format == "JPEG" && strings.EqualFold(ext, ".jpeg")) {
 						if data == nil {
 							data, err = read(actual[key])
 							if err != nil {
@@ -258,9 +267,9 @@ func (e *Editor) compressResourceReferences(parts map[string][]byte, reader *Rea
 							}
 						}
 						stem := strings.TrimSuffix(actual[key], ext)
-						name := stem + ".jpg"
+						name := stem + suffix
 						for n := 1; reserved[name]; n++ {
-							name = fmt.Sprintf("%s-%d.jpg", stem, n)
+							name = fmt.Sprintf("%s-%d%s", stem, n, suffix)
 						}
 						reserved[name] = true
 						locations[key] = name
diff --git a/ofdgo_compression_tiff.go b/ofdgo_compression_tiff.go
new file mode 100644
index 0000000..f16cbcb
--- /dev/null
+++ b/ofdgo_compression_tiff.go
@@ -0,0 +1,167 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"encoding/binary"
+	"image"
+	"math"
+)
+
+// optimizeTIFFResource 压缩单目录TIFF，保留方向、透明度、采样精度及像素密度
+// 入参: ctx 取消上下文, data TIFF编码, options 压缩选项, size 显示像素需求
+// 返回: []byte 图片编码，不可完整转换时保留原文, error 编码或取消错误
+func optimizeTIFFResource(ctx context.Context, data []byte, options CompressionOptions, size image.Point) ([]byte, error) {
+	layout, err := readTIFFLayout(bytes.NewReader(data))
+	if err != nil {
+		return data, ctx.Err()
+	}
+	pixelSize := int64(8)
+	if layout.planar {
+		pixelSize = 16
+	}
+	if int64(layout.width) > outputOptimizationBufferLimit/pixelSize/int64(layout.height) {
+		return data, ctx.Err()
+	}
+	density, ok := tiffCompressionDensity(data, layout)
+	if !ok {
+		return data, ctx.Err()
+	}
+	pixels, err := decodeTIFFImage(ctx, data, layout)
+	if err != nil {
+		return data, ctx.Err()
+	}
+	return optimizeRasterResource(ctx, data, pixels, options, size, density)
+}
+
+// tiffCompressionDensity 校验可转换的TIFF字段并提取规范化后的像素密度
+// 色彩配置、预乘透明度、扩展字段及其他目录需保留原始编码
+// 入参: data TIFF编码, layout 已校验的主图布局
+// 返回: rasterDensity PNG像素密度, bool 是否可完整转换
+func tiffCompressionDensity(data []byte, layout tiffImageLayout) (rasterDensity, bool) {
+	var density rasterDensity
+	var order binary.ByteOrder = binary.LittleEndian
+	if data[0] == 'M' {
+		order = binary.BigEndian
+	}
+	if layout.next != 0 || layout.offset != order.Uint32(data[4:]) || layout.reduced {
+		return density, false
+	}
+	offset := int(layout.offset)
+	count := int(order.Uint16(data[offset:]))
+	unit := uint32(2)
+	var resolutions [2][2]uint32
+	for i := range count {
+		entry := data[offset+2+i*12 : offset+14+i*12]
+		tag, kind, number := order.Uint16(entry), order.Uint16(entry[2:]), order.Uint32(entry[4:])
+		value := order.Uint32(entry[8:])
+		if kind == 3 && number == 1 {
+			value = uint32(order.Uint16(entry[8:]))
+		}
+		switch tag {
+		case 256, 257, 258, 259, 262, 273, 274, 277, 278, 279, 320, 322, 323, 324, 325:
+		case 254:
+			if value != 0 {
+				return density, false
+			}
+		case 255:
+			if value != 1 {
+				return density, false
+			}
+		case 266:
+			if kind != 3 || number != 1 || value != 1 && value != 2 {
+				return density, false
+			}
+		case 284:
+			if kind != 3 || number != 1 || value != 1 && value != 2 {
+				return density, false
+			}
+		case 292, 293:
+			mask := uint32(7)
+			if tag == 293 {
+				mask = 2
+			}
+			if kind != 4 || number != 1 || value & ^mask != 0 {
+				return density, false
+			}
+		case 317:
+			if kind != 3 || number != 1 || value != 1 && value != 2 {
+				return density, false
+			}
+		case 338:
+			if kind != 3 || number != 1 || value != 2 {
+				return density, false
+			}
+		case 339:
+			if kind != 3 || number == 0 || number > 4 {
+				return density, false
+			}
+			values := entry[8:]
+			if number > 2 {
+				if uint64(value)+uint64(number)*2 > uint64(len(data)) {
+					return density, false
+				}
+				values = data[value:]
+			}
+			for j := uint32(0); j < number; j++ {
+				if order.Uint16(values[j*2:]) != 1 {
+					return density, false
+				}
+			}
+		case 282, 283:
+			if kind != 5 || number != 1 || uint64(value)+8 > uint64(len(data)) {
+				return density, false
+			}
+			resolutions[tag-282] = [2]uint32{order.Uint32(data[value:]), order.Uint32(data[value+4:])}
+		case 296:
+			if kind != 3 || number != 1 || value < 1 || value > 3 {
+				return density, false
+			}
+			unit = value
+		default:
+			return density, false
+		}
+	}
+	if resolutions == [2][2]uint32{} {
+		return density, true
+	}
+	x, y := resolutions[0], resolutions[1]
+	if x[0] == 0 || x[1] == 0 || y[0] == 0 || y[1] == 0 {
+		return density, false
+	}
+	if unit == 1 {
+		a, b := uint64(x[0])*uint64(y[1]), uint64(y[0])*uint64(x[1])
+		divisor, next := a, b
+		for next != 0 {
+			divisor, next = next, divisor%next
+		}
+		density.x, density.y = float64(a/divisor), float64(b/divisor)
+	} else {
+		factor := 1 / 0.0254
+		if unit == 3 {
+			factor = 100
+		}
+		density = rasterDensity{x: float64(x[0]) / float64(x[1]) * factor, y: float64(y[0]) / float64(y[1]) * factor, unit: 1}
+	}
+	if math.Round(density.x) < 1 || math.Round(density.y) < 1 || math.Round(density.x) > math.MaxUint32 || math.Round(density.y) > math.MaxUint32 {
+		return density, false
+	}
+	if layout.orientation >= 5 {
+		density.x, density.y = density.y, density.x
+	}
+	return density, true
+}
diff --git a/ofdgo_editor.go b/ofdgo_editor.go
index d37b718..94921bb 100644
--- a/ofdgo_editor.go
+++ b/ofdgo_editor.go
@@ -16,6 +16,7 @@ package ofdgo
 
 import (
 	"bytes"
+	"context"
 	"crypto/rand"
 	"crypto/sha256"
 	"encoding/hex"
@@ -256,6 +257,16 @@ func (e *Editor) CopyPage(index int) (int, error) {
 // 入参: indexes 不重复的原页面索引
 // 返回: []int 副本页面索引, error 错误信息
 func (e *Editor) CopyPages(indexes []int) ([]int, error) {
+	return e.CopyPagesContext(context.Background(), indexes)
+}
+
+// CopyPagesContext 复制页面并响应取消，准备期间不提交页面或撤销记录
+// 入参: ctx 取消上下文, indexes 不重复的原页面索引
+// 返回: []int 副本页面索引, error 校验、复制或取消错误
+func (e *Editor) CopyPagesContext(ctx context.Context, indexes []int) ([]int, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	if err := e.validatePageIndexes(indexes); err != nil {
 		return nil, err
 	}
@@ -263,11 +274,12 @@ func (e *Editor) CopyPages(indexes []int) ([]int, error) {
 		return nil, nil
 	}
 	at := len(e.pages)
-	source, err := e.Reader()
+	source, err := e.readerContext(ctx, nil)
 	if err != nil {
 		return nil, err
 	}
-	if _, err := e.importPages(source, indexes, at, true, PageImportOptions{}); err != nil {
+	defer source.Close()
+	if _, err := e.importPages(ctx, source, indexes, at, true, PageImportOptions{}); err != nil {
 		return nil, err
 	}
 	result := make([]int, len(indexes))
@@ -456,7 +468,7 @@ func (e *Editor) ResizePages(indexes []int, width, height float64) error {
 	return nil
 }
 
-// AddImage 注册PNG、JPEG或JBIG2图片，重复资源复用标识，引用后写入文档
+// AddImage 注册PNG、JPEG、BMP、单页TIFF或JBIG2图片，重复资源复用标识，引用后写入文档
 // 不透明纯黑白PNG仅在无损JBIG2编码更小时转换，其他图片保留原始编码
 // 入参: data 图片数据
 // 返回: string 图片资源标识, error 错误信息
@@ -464,11 +476,18 @@ func (e *Editor) AddImage(data []byte) (string, error) {
 	return e.addDecodedImage(data, nil)
 }
 
+// ImageSize 获取图片资源的像素尺寸，优先复用注册结果，不解码像素
+// 入参: id 图片资源标识
+// 返回: image.Point 像素宽高, error 资源或格式错误
+func (e *Editor) ImageSize(id string) (image.Point, error) {
+	return e.editorImage(id)
+}
+
 // addDecodedImage 注册图片并复用与编码数据一致的像素
 // 入参: data 图片数据, decoded 已解码像素，为nil时按需解码
 // 返回: string 资源标识, error 注册错误
 func (e *Editor) addDecodedImage(data []byte, decoded image.Image) (string, error) {
-	config, format, err := image.DecodeConfig(bytes.NewReader(data))
+	config, format, err := decodeImageConfigData(data)
 	if err != nil {
 		return "", err
 	}
@@ -479,8 +498,8 @@ func (e *Editor) addDecodedImage(data []byte, decoded image.Image) (string, erro
 // 入参: data 图片编码数据, config 图片配置, format 图片格式, decoded 已解码图像，可为nil
 // 返回: string 资源标识, error 格式或资源错误
 func (e *Editor) addImage(data []byte, config image.Config, format string, decoded image.Image) (string, error) {
-	if (format != "png" && format != "jpeg" && format != "jbig2") || config.Width <= 0 || config.Height <= 0 {
-		return "", fmt.Errorf("only PNG, JPEG and JBIG2 images are supported")
+	if (format != "png" && format != "jpeg" && format != "bmp" && format != "tiff" && format != "jbig2") || config.Width <= 0 || config.Height <= 0 {
+		return "", fmt.Errorf("only PNG, JPEG, BMP, single-page TIFF and JBIG2 images are supported")
 	}
 	key := editorResourceKey{checksum: sha256.Sum256(data), index: -1}
 	if id, ok := e.resourceID[key]; ok {
diff --git a/ofdgo_editor_documents.go b/ofdgo_editor_documents.go
index 0052e16..bc22829 100644
--- a/ofdgo_editor_documents.go
+++ b/ofdgo_editor_documents.go
@@ -113,17 +113,15 @@ func (e *Editor) AddDocument(title string, width, height float64, at int) (int,
 		return 0, err
 	}
 	err := e.editDocuments(func(reader *Reader, data []byte, root *editorXML) (int, error) {
+		occupied := make(map[string]bool)
+		for name := range reader.fileNames {
+			directory, _, _ := strings.Cut(name, "/")
+			occupied[directory] = true
+		}
 		directory := ""
 		for number := 0; ; number++ {
 			candidate := fmt.Sprintf("Doc_%d", number)
-			occupied := false
-			for name := range reader.fileNames {
-				if name == candidate || strings.HasPrefix(name, candidate+"/") {
-					occupied = true
-					break
-				}
-			}
-			if !occupied {
+			if !occupied[candidate] {
 				directory = candidate
 				break
 			}
@@ -219,14 +217,25 @@ func (e *Editor) MoveDocument(from, to int) error {
 // 入参: index 文档索引，从0开始
 // 返回: error 删除错误
 func (e *Editor) DeleteDocument(index int) error {
+	return e.DeleteDocumentContext(context.Background(), index)
+}
+
+// DeleteDocumentContext 删除指定文档及其独占文件，取消或失败时保留当前状态和撤销历史
+// 包内至少保留一个文档，其他文档及版本仍在使用的文件不删除
+// 入参: ctx 取消上下文, index 文档索引，从0开始
+// 返回: error 删除或取消错误
+func (e *Editor) DeleteDocumentContext(ctx context.Context, index int) error {
+	if err := ctx.Err(); err != nil {
+		return err
+	}
 	if index < 0 || index >= e.DocumentCount() {
 		return fmt.Errorf("document index %d out of range", index)
 	}
 	if e.DocumentCount() == 1 {
 		return fmt.Errorf("package must contain at least one document")
 	}
-	return e.editDocuments(func(reader *Reader, data []byte, root *editorXML) (int, error) {
-		removed, err := reader.documentFiles(context.Background(), index)
+	return e.editDocumentsContext(ctx, func(reader *Reader, data []byte, root *editorXML) (int, error) {
+		removed, err := reader.documentFiles(ctx, index)
 		if err != nil {
 			return 0, err
 		}
@@ -236,25 +245,35 @@ func (e *Editor) DeleteDocument(index int) error {
 			return name != cleanPackagePath(reader.OFD.DocBody[index].DocRoot) && strings.HasPrefix(name, directory+"/")
 		}) {
 			for name := range reader.fileNames {
+				if err := ctx.Err(); err != nil {
+					return 0, err
+				}
 				if strings.HasPrefix(name, directory+"/") {
 					removed[name] = true
 				}
 			}
 		}
+		var others []int
 		for other := range reader.OFD.DocBody {
-			if other == index {
-				continue
+			if other != index {
+				others = append(others, other)
 			}
-			kept, err := reader.documentFiles(context.Background(), other)
-			if err != nil {
+		}
+		kept, err := reader.documentFiles(ctx, others...)
+		if err != nil {
+			return 0, err
+		}
+		for name := range kept {
+			if err := ctx.Err(); err != nil {
 				return 0, err
 			}
-			for name := range kept {
-				delete(removed, name)
-			}
+			delete(removed, name)
 		}
 		delete(removed, "OFD.xml")
 		for name := range reader.files {
+			if err := ctx.Err(); err != nil {
+				return 0, err
+			}
 			if removed[name] {
 				delete(reader.files, name)
 			}
@@ -280,15 +299,26 @@ func (e *Editor) DeleteDocument(index int) error {
 // 入参: edit 修改包快照并返回新的当前文档索引
 // 返回: error 修改错误
 func (e *Editor) editDocuments(edit func(*Reader, []byte, *editorXML) (int, error)) error {
-	reader, err := e.Reader()
+	ctx := context.Background()
+	if e.output != nil {
+		ctx = e.output.ctx
+	}
+	return e.editDocumentsContext(ctx, edit)
+}
+
+// editDocumentsContext 在独立快照中修改文档入口，确认未取消后提交完整状态
+// 入参: ctx 取消上下文, edit 修改包快照并返回新的当前文档索引
+// 返回: error 修改或取消错误
+func (e *Editor) editDocumentsContext(ctx context.Context, edit func(*Reader, []byte, *editorXML) (int, error)) error {
+	reader, err := e.readerContext(ctx, nil)
 	if err != nil {
 		return err
 	}
-	data, err := reader.readFile("OFD.xml")
+	data, err := reader.readFileViewContext(ctx, "OFD.xml")
 	if err != nil {
 		return err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return err
 	}
@@ -298,6 +328,9 @@ func (e *Editor) editDocuments(edit func(*Reader, []byte, *editorXML) (int, erro
 	if err != nil {
 		return err
 	}
+	if err := ctx.Err(); err != nil {
+		return err
+	}
 	before := e.transactionSnapshot()
 	reader.documentIndex, reader.doc = index, nil
 	if err := reader.initRoot(); err != nil {
@@ -310,9 +343,14 @@ func (e *Editor) editDocuments(edit func(*Reader, []byte, *editorXML) (int, erro
 	if _, err := reader.Doc(); err != nil {
 		return err
 	}
-	if err := e.adoptDocumentReader(reader); err != nil {
+	next := *e
+	if err := next.adoptDocumentReader(reader); err != nil {
 		return err
 	}
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+	*e = next
 	e.recordTransaction(before)
 	return nil
 }
diff --git a/ofdgo_editor_font.go b/ofdgo_editor_font.go
index 2bb66ff..17577af 100644
--- a/ofdgo_editor_font.go
+++ b/ofdgo_editor_font.go
@@ -175,17 +175,30 @@ func editorObjectFontScan(object GraphicObject) bool {
 	return clips != nil || fill != nil && fill.Pattern != nil || stroke != nil && stroke.Pattern != nil
 }
 
-// subsetFonts 收集新增字体实际使用的字形，仅裁剪保存结果，保留完整编辑字体
+// subsetFonts 仅解析实际引用的新增字体，按用字裁剪保存结果，保留完整编辑字体
 // 入参: progress 保存进度回调
 // 返回: map[string][]byte 包内字体子集, error 错误信息
 func (e *Editor) subsetFonts(progress editorProgress) (map[string][]byte, error) {
 	if e.backends.FontResources == nil {
 		return nil, nil
 	}
+	if err := progress.report("fonts", 0, 0); err != nil {
+		return nil, err
+	}
+	if !slices.ContainsFunc(e.resources, func(resource editorResource) bool {
+		return resource.font != nil && resource.font.FontFile != ""
+	}) {
+		return nil, nil
+	}
+	fonts, _, _, files := e.usedResources()
+	total := len(fonts) + len(e.pages) + 2*len(e.resources)
 	used := make(map[string]map[uint16]bool)
 	mapped := make(map[string]bool)
-	for _, resource := range e.resources {
-		if resource.font == nil || resource.font.FontFile == "" {
+	for i, resource := range fonts {
+		if err := progress.report("fonts", i, total); err != nil {
+			return nil, err
+		}
+		if resource.font.FontFile == "" {
 			continue
 		}
 		resourceFont := e.fonts[resource.font.ID]
@@ -202,7 +215,10 @@ func (e *Editor) subsetFonts(progress editorProgress) (map[string][]byte, error)
 		}
 	}
 	if len(used) == 0 {
-		return nil, progress.report("fonts", 0, 0)
+		for i := range e.resources {
+			e.resources[i].subset = nil
+		}
+		return nil, progress.report("fonts", total, total)
 	}
 	refs := editorResourceRefs{ids: make(map[string]bool), files: make(map[string]bool)}
 	nested := false
@@ -222,9 +238,8 @@ func (e *Editor) subsetFonts(progress editorProgress) (map[string][]byte, error)
 			}
 		}
 	}
-	total := len(e.pages) + len(e.resources)
 	for i, page := range e.pages {
-		if err := progress.report("fonts", i, total); err != nil {
+		if err := progress.report("fonts", len(fonts)+i, total); err != nil {
 			return nil, err
 		}
 		for _, layer := range page.Content.Layer {
@@ -278,12 +293,11 @@ func (e *Editor) subsetFonts(progress editorProgress) (map[string][]byte, error)
 		}
 	}
 	if nested {
-		_, _, _, files := e.usedResources()
 		for i, resource := range e.resources {
-			if resource.definition() == "" || !slices.Contains(files, resource.name) {
+			if resource.definition() == "" || !files[resource.name] {
 				continue
 			}
-			if err := progress.report("fonts", len(e.pages)+i, total); err != nil {
+			if err := progress.report("fonts", len(fonts)+len(e.pages)+i, total); err != nil {
 				return nil, err
 			}
 			if resource.usage.matches(resource.name, resource.data) {
@@ -315,7 +329,7 @@ func (e *Editor) subsetFonts(progress editorProgress) (map[string][]byte, error)
 	}
 	result := make(map[string][]byte)
 	for i := range e.resources {
-		if err := progress.report("fonts", len(e.pages)+i, total); err != nil {
+		if err := progress.report("fonts", len(fonts)+len(e.pages)+len(e.resources)+i, total); err != nil {
 			return nil, err
 		}
 		resource := &e.resources[i]
diff --git a/ofdgo_editor_images.go b/ofdgo_editor_images.go
index 0a0b590..33c0b15 100644
--- a/ofdgo_editor_images.go
+++ b/ofdgo_editor_images.go
@@ -40,7 +40,7 @@ type ImagePageOptions struct {
 	OnProgress    func(stage string, completed, total int) error
 }
 
-// ImportImages 将PNG、JPEG或JBIG2逐张建页并插入指定位置，一次操作计入一条撤销记录
+// ImportImages 将PNG、JPEG、BMP、单页TIFF或JBIG2逐张建页并插入指定位置，一次操作计入一条撤销记录
 // 保留像素及透明度，方向通过对象变换实现，不重新采样；图片压缩沿用AddImage规则
 // 失败或取消时不修改文档，读取期间仍需容纳单张图片的解码数据
 // 入参: ctx 取消上下文, sources 按页面顺序排列的图片, at 插入位置, options 建页选项
@@ -110,14 +110,19 @@ func (e *Editor) ImportImages(ctx context.Context, sources []ImagePageSource, at
 // 入参: data 图片数据, options 建页选项
 // 返回: int 从0开始的页面索引, error 错误信息
 func (e *Editor) addImagePage(data []byte, options ImagePageOptions) (int, error) {
-	config, format, err := image.DecodeConfig(bytes.NewReader(data))
+	config, format, err := decodeImageConfigData(data)
 	if err != nil {
 		return 0, err
 	}
-	if format != "png" && format != "jpeg" && format != "jbig2" {
-		return 0, fmt.Errorf("only PNG, JPEG and JBIG2 images are supported")
+	if format != "png" && format != "jpeg" && format != "bmp" && format != "tiff" && format != "jbig2" {
+		return 0, fmt.Errorf("only PNG, JPEG, BMP, single-page TIFF and JBIG2 images are supported")
+	}
+	var decoded image.Image
+	if format == "bmp" || format == "tiff" {
+		decoded, _, err = decodeImageData(data)
+	} else {
+		decoded, _, err = image.Decode(bytes.NewReader(data))
 	}
-	decoded, _, err := image.Decode(bytes.NewReader(data))
 	if err != nil {
 		return 0, err
 	}
diff --git a/ofdgo_editor_import.go b/ofdgo_editor_import.go
index e01a138..957a16a 100644
--- a/ofdgo_editor_import.go
+++ b/ofdgo_editor_import.go
@@ -32,15 +32,19 @@ type editorImportEntry struct {
 	name, base, group string
 	data              []byte
 	node              *editorXML
+	template          bool
+	ids               map[string]string
 }
 
 // editorPageImport 一次插页操作的独立资源图，提交前不修改目标文档
 type editorPageImport struct {
+	ctx           context.Context
 	reader        *Reader
 	doc           *Document
 	directory     string
 	target        *Reader
 	paths         map[string]string
+	templatePaths map[string]string
 	evidence      map[string]string
 	resourceXML   []byte
 	pageParts     []byte
@@ -95,6 +99,17 @@ type editorProgressReader struct {
 // 入参: source 来源编辑器, sourcePage 来源页, ids 对象标识, page 目标页, options 迁移选项
 // 返回: []string 新对象标识, error 错误信息
 func (e *Editor) ImportObjects(source *Editor, sourcePage int, ids []string, page int, options ObjectImportOptions) ([]string, error) {
+	return e.ImportObjectsContext(context.Background(), source, sourcePage, ids, page, options)
+}
+
+// ImportObjectsContext 迁移对象及资源，取消时不提交修改或撤销记录
+// 入参: ctx 取消上下文, source 来源编辑器, sourcePage 来源页, ids 对象标识, page 目标页, options 迁移选项
+// 返回: []string 新对象标识, error 校验、迁移或取消错误
+func (e *Editor) ImportObjectsContext(ctx context.Context, source *Editor, sourcePage int, ids []string, page int, options ObjectImportOptions) ([]string, error) {
+	progress := editorProgress(options.OnProgress).withContext(ctx)
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	if !finite(options.DX) || !finite(options.DY) {
 		return nil, fmt.Errorf("import requires finite offsets")
 	}
@@ -110,19 +125,22 @@ func (e *Editor) ImportObjects(source *Editor, sourcePage int, ids []string, pag
 		return nil, err
 	}
 	if source == e {
-		return e.CopyObjects(page, objects, options.DX, options.DY)
+		if err := progress.report("commit", 0, 0); err != nil {
+			return nil, err
+		}
+		return e.CopyObjectsContext(ctx, page, objects, options.DX, options.DY)
 	}
-	reader, err := source.Reader()
+	reader, err := source.readerContext(ctx, progress)
 	if err != nil {
 		return nil, err
 	}
+	defer reader.Close()
 	doc, err := reader.Doc()
 	if err != nil {
 		return nil, err
 	}
 	ref := doc.Pages.Page[sourcePage]
-	progress := editorProgress(options.OnProgress)
-	base, migration, err := e.prepareImport(reader, doc, map[string]bool{ref.ID: true}, false, progress)
+	base, migration, err := e.prepareImport(ctx, reader, doc, map[string]bool{ref.ID: true}, false, progress)
 	if err != nil {
 		return nil, err
 	}
@@ -169,7 +187,7 @@ func (e *Editor) ImportObjects(source *Editor, sourcePage int, ids []string, pag
 		if err != nil {
 			return nil, err
 		}
-		root, err := parseEditorXML(data)
+		root, err := parseEditorXMLContext(ctx, data)
 		if err != nil {
 			return nil, err
 		}
@@ -207,7 +225,7 @@ func (e *Editor) ImportObjects(source *Editor, sourcePage int, ids []string, pag
 				return err
 			}
 			layer := content.Layer[0]
-			root, err := parseEditorXML(data)
+			root, err := parseEditorXMLContext(ctx, data)
 			if err != nil {
 				return err
 			}
@@ -221,7 +239,7 @@ func (e *Editor) ImportObjects(source *Editor, sourcePage int, ids []string, pag
 			imported[i].CompositeGraphicUnit.states = remapCompositeStates(objects[i].CompositeGraphicUnit.states, migration.ids)
 			imported[i].origin = &editorObjectOrigin{editor: editor, document: editor.documentIdentity(), data: data, node: root.child("Layer").children[0], object: object}
 		}
-		copied, err = editor.CopyObjects(page, imported, options.DX, options.DY)
+		copied, err = editor.CopyObjectsContext(ctx, page, imported, options.DX, options.DY)
 		return err
 	})
 	return copied, err
@@ -244,10 +262,21 @@ func (e *Editor) ImportPages(source *Reader, indexes []int, at int) ([]string, e
 // 入参: source 来源阅读器, indexes 来源页面索引, at 目标插入位置, options 导入选项
 // 返回: []string 新页面标识, error 错误信息
 func (e *Editor) ImportPagesWithOptions(source *Reader, indexes []int, at int, options PageImportOptions) ([]string, error) {
+	return e.ImportPagesContext(context.Background(), source, indexes, at, options)
+}
+
+// ImportPagesContext 导入页面及关联资源，取消检查覆盖共享资源隔离和版本清单更新
+// 取消时不提交页面、目录或撤销记录，其他行为与ImportPagesWithOptions相同
+// 入参: ctx 取消上下文, source 来源阅读器, indexes 来源页索引, at 插入位置, options 导入选项
+// 返回: []string 新页面标识, error 校验、导入或取消错误
+func (e *Editor) ImportPagesContext(ctx context.Context, source *Reader, indexes []int, at int, options PageImportOptions) ([]string, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	if source.encryption != nil && e.encryption == nil && !options.AllowDecrypted {
 		return nil, ErrEncryptionPolicyRequired
 	}
-	return e.importPages(source, indexes, at, false, options)
+	return e.importPages(ctx, source, indexes, at, false, options)
 }
 
 // Read 读取数据并在下一块开始前检查取消
@@ -263,9 +292,12 @@ func (r *editorProgressReader) Read(data []byte) (int, error) {
 }
 
 // importPages 迁移页面，同文档复制时复用原资源并保留其他页面的跳转
-// 入参: source 来源阅读器, indexes 来源页索引, at 插入位置, copyPage 是否同文档复制, options 导入选项
+// 入参: ctx 取消上下文, source 来源阅读器, indexes 来源页索引, at 插入位置, copyPage 是否同文档复制, options 导入选项
 // 返回: []string 新页标识, error 错误信息
-func (e *Editor) importPages(source *Reader, indexes []int, at int, copyPage bool, options PageImportOptions) ([]string, error) {
+func (e *Editor) importPages(ctx context.Context, source *Reader, indexes []int, at int, copyPage bool, options PageImportOptions) ([]string, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	if at < 0 || at > len(e.pages) {
 		return nil, fmt.Errorf("page index %d out of range", at)
 	}
@@ -287,8 +319,8 @@ func (e *Editor) importPages(source *Reader, indexes []int, at int, copyPage boo
 	if len(indexes) == 0 {
 		return nil, nil
 	}
-	progress := editorProgress(options.OnProgress)
-	base, m, err := e.prepareImport(source, doc, selected, copyPage, progress)
+	progress := editorProgress(options.OnProgress).withContext(ctx)
+	base, m, err := e.prepareImport(ctx, source, doc, selected, copyPage, progress)
 	if err != nil {
 		return nil, err
 	}
@@ -300,7 +332,7 @@ func (e *Editor) importPages(source *Reader, indexes []int, at int, copyPage boo
 		}
 		page := doc.Pages.Page[index]
 		ids[i] = m.id(page.ID)
-		name, err := m.copyXML(source.ResPath(page.BaseLoc), packagePagePath(m.directory, ids[i]))
+		name, err := m.copyXML(source.ResPath(page.BaseLoc), packagePagePath(m.directory, ids[i]), false)
 		if err != nil {
 			return nil, err
 		}
@@ -328,7 +360,7 @@ func (e *Editor) importPages(source *Reader, indexes []int, at int, copyPage boo
 		if err != nil {
 			return nil, err
 		}
-		outlines, err = e.appendImportedOutlines(imported)
+		outlines, err = e.appendImportedOutlines(ctx, imported)
 		if err != nil {
 			return nil, err
 		}
@@ -353,6 +385,9 @@ func (e *Editor) importPages(source *Reader, indexes []int, at int, copyPage boo
 		preview := *e
 		preview.source, preview.pages = next, added
 		for i, index := range indexes {
+			if err := ctx.Err(); err != nil {
+				return nil, err
+			}
 			states := make(map[string]editorCompositeState)
 			groups := make(map[string]map[string]editorCompositeState)
 			for _, layer := range e.pages[index].Content.Layer {
@@ -433,9 +468,9 @@ func (e *Editor) bindImportResources() {
 }
 
 // prepareImport 建立页面和对象迁移共用的资源图，不修改目标编辑器
-// 入参: source 来源, doc 来源文档, selected 页面映射范围, copyPage 是否同文档, progress 进度
+// 入参: ctx 取消上下文, source 来源, doc 来源文档, selected 页面映射范围, copyPage 是否同文档, progress 进度
 // 返回: *editorSource 目标基础, *editorPageImport 资源图, error 错误信息
-func (e *Editor) prepareImport(source *Reader, doc *Document, selected map[string]bool, copyPage bool, progress editorProgress) (*editorSource, *editorPageImport, error) {
+func (e *Editor) prepareImport(ctx context.Context, source *Reader, doc *Document, selected map[string]bool, copyPage bool, progress editorProgress) (*editorSource, *editorPageImport, error) {
 	maximum, err := e.sourceMaxID(progress)
 	if err != nil {
 		return nil, nil, err
@@ -447,7 +482,7 @@ func (e *Editor) prepareImport(source *Reader, doc *Document, selected map[strin
 			return nil, nil, err
 		}
 	}
-	m := &editorPageImport{reader: source, doc: doc, directory: e.packageDirectory(), target: base.reader, maximum: maximum, copyPage: copyPage, progress: progress,
+	m := &editorPageImport{ctx: ctx, reader: source, doc: doc, directory: e.packageDirectory(), target: base.reader, maximum: maximum, copyPage: copyPage, progress: progress,
 		paths: make(map[string]string), evidence: make(map[string]string), ids: make(map[string]string), pages: make(map[string]bool, len(selected)),
 		files: make(map[string][]byte), resources: make(map[string]editorImportEntry), used: make(map[string][]byte),
 		templates: make(map[string]TemplatePage), templateParts: make(map[string][]byte), attachments: make(map[string][]byte)}
@@ -461,7 +496,7 @@ func (e *Editor) prepareImport(source *Reader, doc *Document, selected map[strin
 	if err != nil {
 		return nil, nil, err
 	}
-	m.documentRoot, err = parseEditorXML(m.documentXML)
+	m.documentRoot, err = parseEditorXMLContext(ctx, m.documentXML)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -501,6 +536,9 @@ func (m *editorPageImport) encodeResources() error {
 	for _, group := range []string{"ColorSpaces", "DrawParams", "Fonts", "MultiMedias", "CompositeGraphicUnits"} {
 		var entries []byte
 		for _, id := range ids {
+			if err := m.ctx.Err(); err != nil {
+				return err
+			}
 			if m.resources[id].group == group {
 				entries = append(entries, m.used[id]...)
 			}
@@ -542,7 +580,7 @@ func (m *editorPageImport) outlines() ([]byte, error) {
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -591,7 +629,7 @@ func (m *editorPageImport) outlines() ([]byte, error) {
 		}
 	}
 	data = editorPatchXML(data, patches)
-	root, err = parseEditorXML(data)
+	root, err = parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -624,13 +662,13 @@ func (m *editorPageImport) outlines() ([]byte, error) {
 }
 
 // appendImportedOutlines 将已迁移目录追加到目标根目录，规范计数且不修改当前编辑器
-// 入参: imported 已迁移目录片段
+// 入参: ctx 取消上下文, imported 已迁移目录片段
 // 返回: []byte 合并后的目录, error 错误信息
-func (e *Editor) appendImportedOutlines(imported []byte) ([]byte, error) {
+func (e *Editor) appendImportedOutlines(ctx context.Context, imported []byte) ([]byte, error) {
 	if len(imported) == 0 {
 		return e.outlines, nil
 	}
-	added, err := parseEditorXML(imported)
+	added, err := parseEditorXMLContext(ctx, imported)
 	if err != nil {
 		return nil, err
 	}
@@ -638,7 +676,7 @@ func (e *Editor) appendImportedOutlines(imported []byte) ([]byte, error) {
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -647,6 +685,9 @@ func (e *Editor) appendImportedOutlines(imported []byte) ([]byte, error) {
 		content = bytes.Clone(data[root.open:root.close])
 	}
 	for _, child := range editorOutlineChildren(added) {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		value, err := editorXMLStandalone(imported[child.start:child.end], child)
 		if err != nil {
 			return nil, err
@@ -732,7 +773,7 @@ func (m *editorPageImport) resourceIndex(name string) error {
 	if err != nil {
 		return err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return err
 	}
@@ -773,7 +814,7 @@ func (m *editorPageImport) reference(value string) (string, error) {
 	if template, ok := m.templates[key]; ok {
 		if _, seen := m.templateParts[key]; !seen {
 			m.templateParts[key] = nil
-			name, err := m.copyXML(m.reader.ResPath(template.BaseLoc), path.Join(m.directory, "Templates", "Template_"+m.id(value), "Content.xml"))
+			name, err := m.copyXML(m.reader.ResPath(template.BaseLoc), path.Join(m.directory, "Templates", "Template_"+m.id(value), "Content.xml"), true)
 			if err != nil {
 				return "", err
 			}
@@ -801,7 +842,7 @@ func (m *editorPageImport) documentReference(group, kind, source, id, loc string
 				return nil, err
 			}
 			for _, attr := range [][2]string{{"ID", id}, {"BaseLoc", loc}} {
-				root, err := parseEditorXML(data)
+				root, err := parseEditorXMLContext(m.ctx, data)
 				if err != nil {
 					return nil, err
 				}
@@ -857,11 +898,18 @@ func (m *editorPageImport) copyData(name, relative string, evidence bool) (strin
 }
 
 // copyXML 复制页面或标准关联XML并重映射引用
-// 入参: name 来源路径, candidate 目标候选路径
+// 入参: name 来源路径, candidate 目标候选路径, template 是否按模板页复制
 // 返回: string 新路径, error 错误信息
-func (m *editorPageImport) copyXML(name, candidate string) (string, error) {
+func (m *editorPageImport) copyXML(name, candidate string, template bool) (string, error) {
+	paths := m.paths
+	if template {
+		if m.templatePaths == nil {
+			m.templatePaths = make(map[string]string)
+		}
+		paths = m.templatePaths
+	}
 	key := cleanPackagePath(name)
-	if target := m.paths[key]; target != "" {
+	if target := paths[key]; target != "" {
 		return "/" + target, nil
 	}
 	target := packageAvailableName(m.target, m.files, candidate)
@@ -869,11 +917,11 @@ func (m *editorPageImport) copyXML(name, candidate string) (string, error) {
 	if err != nil {
 		return "", err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return "", err
 	}
-	m.paths[key], m.files[target] = target, nil
+	paths[key], m.files[target] = target, nil
 	if root.name.Local == "Page" {
 		for _, child := range root.children {
 			if child.name.Local == "PageRes" {
@@ -887,7 +935,14 @@ func (m *editorPageImport) copyXML(name, candidate string) (string, error) {
 			}
 		}
 	}
-	encoded, err := m.encode(editorImportEntry{name: name, data: data}, root)
+	entry := editorImportEntry{name: name, data: data, template: template}
+	if template && slices.ContainsFunc(m.doc.Pages.Page, func(page Page) bool {
+		return cleanPackagePath(m.reader.ResPath(page.BaseLoc)) == key
+	}) {
+		entry.ids = make(map[string]string)
+		m.templateObjectIDs(root, entry.ids)
+	}
+	encoded, err := m.encode(entry, root)
 	if err != nil {
 		return "", err
 	}
@@ -895,6 +950,21 @@ func (m *editorPageImport) copyXML(name, candidate string) (string, error) {
 	return "/" + target, nil
 }
 
+// templateObjectIDs 为兼作普通页的模板内容分配独立对象标识
+// 入参: node 当前节点, ids 模板副本的标识映射
+func (m *editorPageImport) templateObjectIDs(node *editorXML, ids map[string]string) {
+	if value := node.attr("ID"); value != "" {
+		key := editorImportKey(value)
+		if ids[key] == "" {
+			m.maximum++
+			ids[key] = strconv.Itoa(m.maximum)
+		}
+	}
+	for _, child := range node.children {
+		m.templateObjectIDs(child, ids)
+	}
+}
+
 // editorImportText 获取节点直接文本，解码XML实体
 // 入参: data XML, node 节点
 // 返回: string 文本
@@ -910,6 +980,11 @@ func editorImportText(data []byte, node *editorXML) string {
 // 入参: entry 来源上下文, node 当前节点
 // 返回: []byte 独立XML片段, error 错误信息
 func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]byte, error) {
+	if m.ctx != nil {
+		if err := m.ctx.Err(); err != nil {
+			return nil, err
+		}
+	}
 	if classifyOFDNamespace(node.name.Space) == ofdXMLUnknown {
 		return nil, fmt.Errorf("cannot remap extension namespace %q", node.name.Space)
 	}
@@ -961,7 +1036,9 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 		var err error
 		switch key {
 		case "ID":
-			if name == "Attachment" {
+			if id := entry.ids[editorImportKey(value)]; id != "" {
+				value = id
+			} else if name == "Attachment" {
 				value, err = m.attachmentID(value)
 			} else if name == "Signature" || name == "StampAnnot" {
 				value, err = m.signatureID(entry.name, name, value)
@@ -992,7 +1069,7 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 					}
 					directory = path.Join("Signs", "Sign_"+id)
 				}
-				value, err = m.copyXML(loc, path.Join(m.directory, directory, path.Base(loc)))
+				value, err = m.copyXML(loc, path.Join(m.directory, directory, path.Base(loc)), false)
 			}
 		case "AttachID":
 			value, err = m.attachment(value)
@@ -1032,15 +1109,24 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 	var content []byte
 	pageContent := name == "Page" && node.parent == nil
 	if pageContent {
-		area := m.doc.CommonData.PageArea
 		local := node.child("Area")
-		var fields []byte
-		for _, field := range [][2]string{{"PhysicalBox", area.PhysicalBox}, {"ApplicationBox", area.ApplicationBox}, {"ContentBox", area.ContentBox}, {"BleedBox", area.BleedBox}} {
-			if field[1] != "" && (local == nil || local.child(field[0]) == nil) {
-				fields = append(fields, editorXMLText(field[0], field[1])...)
-			}
-		}
 		if local == nil {
+			var templates []Template
+			for _, child := range node.children {
+				if !entry.template && child.matchesOFD("Template") {
+					templates = append(templates, Template{TemplateID: child.attr("TemplateID")})
+				}
+			}
+			area, err := m.reader.resolvePageArea(PageArea{}, templates)
+			if err != nil {
+				return nil, err
+			}
+			var fields []byte
+			for _, field := range [][2]string{{"PhysicalBox", area.PhysicalBox}, {"ApplicationBox", area.ApplicationBox}, {"ContentBox", area.ContentBox}, {"BleedBox", area.BleedBox}} {
+				if field[1] != "" {
+					fields = append(fields, editorXMLText(field[0], field[1])...)
+				}
+			}
 			encoded, err := editorXMLContainer("Area", nil, fields)
 			if err != nil {
 				return nil, err
@@ -1051,17 +1137,13 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 			if err != nil {
 				return nil, err
 			}
-			root, err := parseEditorXML(encoded)
-			if err != nil {
-				return nil, err
-			}
-			if root.open != root.end {
-				fields = append(bytes.Clone(encoded[root.open:root.close]), fields...)
-			}
-			content = editorPatchXML(encoded, []editorXMLPatch{editorXMLContent(encoded, root, fields)})
+			content = encoded
 		}
 	}
 	for _, child := range node.children {
+		if pageContent && entry.template && child.matchesOFD("Template") {
+			continue
+		}
 		if pageContent && child.name.Local == "Area" {
 			continue
 		}
@@ -1075,7 +1157,7 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 				if err != nil {
 					return nil, err
 				}
-				root, err := parseEditorXML(encoded)
+				root, err := parseEditorXMLContext(m.ctx, encoded)
 				if err != nil {
 					return nil, err
 				}
@@ -1133,7 +1215,7 @@ func (m *editorPageImport) encode(entry editorImportEntry, node *editorXML) ([]b
 				value, err = m.copyData(loc, path.Join("Attachments", path.Base(loc)), false)
 			} else {
 				candidate := path.Join(m.directory, "Annotations", "Page_"+m.id(node.parent.attr("PageID"))+".xml")
-				value, err = m.copyXML(loc, candidate)
+				value, err = m.copyXML(loc, candidate, false)
 			}
 		case "Font", "Substitution", "Thumbnail":
 			value, err = m.reference(strings.TrimSpace(value))
@@ -1162,7 +1244,7 @@ func (m *editorPageImport) annotations() ([]byte, error) {
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -1198,7 +1280,7 @@ func (m *editorPageImport) attachment(id string) (string, error) {
 	if err != nil {
 		return "", err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return "", err
 	}
@@ -1261,7 +1343,7 @@ func (m *editorPageImport) signatureID(source, kind, id string) (string, error)
 		}
 		if doc.Signatures != "" {
 			name := m.target.ResPath(doc.Signatures)
-			data, err := m.target.readFile(name)
+			data, err := m.target.readFileViewContext(m.ctx, name)
 			if err != nil {
 				return "", err
 			}
@@ -1272,7 +1354,7 @@ func (m *editorPageImport) signatureID(source, kind, id string) (string, error)
 			sequence.observe(list.MaxSignID)
 			for _, ref := range list.List {
 				sequence.observe(ref.ID)
-				data, err := m.target.readFile(signatureRefPath(name, ref.BaseLoc))
+				data, err := m.target.readFileViewContext(m.ctx, signatureRefPath(name, ref.BaseLoc))
 				if err != nil {
 					return "", err
 				}
@@ -1306,7 +1388,7 @@ func (m *editorPageImport) signatures() ([]byte, error) {
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -1347,6 +1429,9 @@ func (m *editorPageImport) signatures() ([]byte, error) {
 // 入参: base 目标基底, refs 插入页面, annotations 注释索引, signatures 签名索引
 // 返回: *editorSource 新基底, error 错误信息
 func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, signatures []byte) (*editorSource, error) {
+	if err := m.ctx.Err(); err != nil {
+		return nil, err
+	}
 	files := maps.Clone(base.reader.files)
 	if files == nil {
 		files = make(map[string][]byte)
@@ -1356,11 +1441,11 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 	if file, ok := base.reader.packageFile(name); ok {
 		name = cleanPackagePath(file.Name)
 	}
-	data, err := base.reader.readFile(name)
+	data, err := base.reader.readFileViewContext(m.ctx, name)
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(m.ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -1415,12 +1500,15 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 	}
 	data = editorPatchXML(data, patches)
 	if len(annotations) != 0 {
-		loc, err := mergeImportIndex(base.reader, files, base.document.Annotations, path.Join(m.directory, "Annotations.xml"), "Annotations", annotations)
+		loc, err := mergeImportIndex(m.ctx, base.reader, files, base.document.Annotations, path.Join(m.directory, "Annotations.xml"), "Annotations", annotations)
 		if err != nil {
 			return nil, err
 		}
 		if base.document.Annotations == "" {
-			root, _ = parseEditorXML(data)
+			root, err = parseEditorXMLContext(m.ctx, data)
+			if err != nil {
+				return nil, err
+			}
 			data, err = editorXMLSetText(data, root, [][2]string{{"Annotations", loc}})
 			if err != nil {
 				return nil, err
@@ -1432,7 +1520,10 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 		for _, id := range slices.Sorted(maps.Keys(m.attachments)) {
 			entries = append(entries, m.attachments[id]...)
 		}
-		root, _ = parseEditorXML(data)
+		root, err = parseEditorXMLContext(m.ctx, data)
+		if err != nil {
+			return nil, err
+		}
 		node := root.child("Attachments")
 		if node != nil && len(node.children) != 0 {
 			entries, err = editorXMLGenerated(entries, node.name.Space)
@@ -1442,7 +1533,7 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 			content := append(bytes.Clone(data[node.open:node.close]), entries...)
 			data = editorPatchXML(data, []editorXMLPatch{editorXMLContent(data, node, content)})
 		} else {
-			loc, err := mergeImportIndex(base.reader, files, base.document.Attachments.Path, path.Join(m.directory, "Attachments.xml"), "Attachments", entries)
+			loc, err := mergeImportIndex(m.ctx, base.reader, files, base.document.Attachments.Path, path.Join(m.directory, "Attachments.xml"), "Attachments", entries)
 			if err != nil {
 				return nil, err
 			}
@@ -1459,13 +1550,13 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 		if base.reader.OFD.DocBody[base.reader.documentIndex].versioned {
 			return nil, fmt.Errorf("signature import into a versioned document is not supported")
 		}
-		loc, err := mergeImportIndex(base.reader, files, base.document.Signatures, path.Join(m.directory, "Signs", "Signatures.xml"), "Signatures", signatures)
+		loc, err := mergeImportIndex(m.ctx, base.reader, files, base.document.Signatures, path.Join(m.directory, "Signs", "Signatures.xml"), "Signatures", signatures)
 		if err != nil {
 			return nil, err
 		}
 		indexName := cleanPackagePath(loc)
 		indexData := files[indexName]
-		indexRoot, err := parseEditorXML(indexData)
+		indexRoot, err := parseEditorXMLContext(m.ctx, indexData)
 		if err != nil {
 			return nil, err
 		}
@@ -1477,11 +1568,11 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 		}
 		files[indexName] = editorPatchXML(indexData, []editorXMLPatch{patch})
 		if base.document.Signatures == "" {
-			ofd, err := base.reader.readFile("OFD.xml")
+			ofd, err := base.reader.readFileViewContext(m.ctx, "OFD.xml")
 			if err != nil {
 				return nil, err
 			}
-			root, err := parseEditorXML(ofd)
+			root, err := parseEditorXMLContext(m.ctx, ofd)
 			if err != nil {
 				return nil, err
 			}
@@ -1495,7 +1586,7 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 			}
 		}
 	}
-	parts, err := base.reader.versionChanges(context.Background(), files)
+	parts, err := base.reader.versionChanges(m.ctx, files)
 	if err != nil {
 		return nil, err
 	}
@@ -1528,9 +1619,9 @@ func (m *editorPageImport) merge(base *editorSource, refs []Page, annotations, s
 }
 
 // mergeImportIndex 向原索引追加条目，不移动原条目的相对路径基准
-// 入参: reader 目标输入包, files 新条目, loc 原索引, fallback 新索引路径, rootName 根节点名称, entries 新条目
+// 入参: ctx 取消上下文, reader 目标输入包, files 新条目, loc 原索引, fallback 新索引路径, rootName 根节点名称, entries 新条目
 // 返回: string 索引绝对路径, error 错误信息
-func mergeImportIndex(reader *Reader, files map[string][]byte, loc, fallback, rootName string, entries []byte) (string, error) {
+func mergeImportIndex(ctx context.Context, reader *Reader, files map[string][]byte, loc, fallback, rootName string, entries []byte) (string, error) {
 	name := fallback
 	var data []byte
 	var err error
@@ -1545,12 +1636,12 @@ func mergeImportIndex(reader *Reader, files map[string][]byte, loc, fallback, ro
 		var exists bool
 		data, exists = files[name]
 		if !exists {
-			data, err = reader.readFile(name)
+			data, err = reader.readFileViewContext(ctx, name)
 		}
 		if err != nil {
 			return "", err
 		}
-		root, parseErr := parseEditorXML(data)
+		root, parseErr := parseEditorXMLContext(ctx, data)
 		if parseErr != nil {
 			return "", parseErr
 		}
diff --git a/ofdgo_editor_open.go b/ofdgo_editor_open.go
index 5aa4246..32fe79d 100644
--- a/ofdgo_editor_open.go
+++ b/ofdgo_editor_open.go
@@ -912,7 +912,14 @@ func (e *Editor) editorImage(id string) (image.Point, error) {
 		return image.Point{}, err
 	}
 	defer input.Close()
-	config, _, err := image.DecodeConfig(input)
+	config, _, err := decodeImageConfig(input)
+	if errors.Is(err, errTIFFRandomAccess) {
+		data, readErr := reader.readFileView(reader.ResPath(name))
+		if readErr != nil {
+			return image.Point{}, readErr
+		}
+		config, _, err = decodeImageConfigData(data)
+	}
 	if err != nil {
 		return image.Point{}, err
 	}
diff --git a/ofdgo_editor_package.go b/ofdgo_editor_package.go
index b696060..5cafde4 100644
--- a/ofdgo_editor_package.go
+++ b/ofdgo_editor_package.go
@@ -29,14 +29,14 @@ import (
 )
 
 // sourceParts 生成编辑快照的XML和新增资源，删页时同步清理标准引用
-// 入参: progress 保存进度回调，预览时为nil
+// 入参: ctx 取消上下文, progress 保存进度回调，预览时为nil
 // 返回: map[string][]byte 替换及新增条目, error 错误信息
-func (e *Editor) sourceParts(progress editorProgress) (map[string][]byte, error) {
+func (e *Editor) sourceParts(ctx context.Context, progress editorProgress) (map[string][]byte, error) {
 	parts, err := e.sourcePartsReferences(progress, nil)
 	if err != nil {
 		return nil, err
 	}
-	return e.source.reader.versionChanges(context.Background(), parts)
+	return e.source.reader.versionChanges(ctx, parts)
 }
 
 // sourcePartsReferences 生成保存快照，同步收集完全自产页面的实际编码引用
@@ -119,18 +119,17 @@ func (e *Editor) sourcePartsPrepared(progress editorProgress, generated map[stri
 	if err := progress.report("pages", len(e.pages), len(e.pages)); err != nil {
 		return nil, err
 	}
-	fonts, images, spaces, resourceFiles := e.usedResources()
+	fonts, images, spaces, usedFiles := e.usedResources()
 	for _, resource := range e.resources {
-		if resource.definition() != "" && slices.Contains(resourceFiles, resource.name) || resource.colorProfileUsed(spaces) {
+		if resource.definition() != "" && usedFiles[resource.name] || resource.colorProfileUsed(spaces) {
 			parts[resource.name] = resource.data
 		}
 	}
 	commonData := source.document.CommonData
 	for _, name := range append(slices.Clone(commonData.DocumentRes), commonData.PublicRes...) {
-		resourceFiles = slices.DeleteFunc(resourceFiles, func(file string) bool {
-			return file == reader.ResPath(name)
-		})
+		delete(usedFiles, reader.ResPath(name))
 	}
+	resourceFiles := slices.Sorted(maps.Keys(usedFiles))
 	if len(fonts)+len(images)+len(spaces) != 0 {
 		data, err := e.encodeXML(func(x *ofdXML) { x.resources(fonts, images, spaces) })
 		if err != nil {
@@ -720,10 +719,10 @@ func (e *Editor) writeSource(writer io.Writer, fonts map[string][]byte, progress
 }
 
 // sourceReader 生成原包与修改条目组成的独立预览快照，不进行ZIP压缩
-// 入参: progress 准备进度回调
+// 入参: ctx 取消上下文, progress 准备进度回调
 // 返回: *Reader 阅读器, error 错误信息
-func (e *Editor) sourceReader(progress editorProgress) (*Reader, error) {
-	parts, err := e.sourceParts(progress)
+func (e *Editor) sourceReader(ctx context.Context, progress editorProgress) (*Reader, error) {
+	parts, err := e.sourceParts(ctx, progress)
 	if err != nil {
 		return nil, err
 	}
diff --git a/ofdgo_editor_reference.go b/ofdgo_editor_reference.go
index e895ced..dc9af6c 100644
--- a/ofdgo_editor_reference.go
+++ b/ofdgo_editor_reference.go
@@ -193,8 +193,6 @@ func (s *editorReferenceScan) accept(token xml.Token) bool {
 				if token.Name.Local != "TextObject" && token.Name.Local != "Text" {
 					usage.unsafe = true
 				}
-			} else if attr.Name.Local == "Substitution" {
-				s.refs.fontUsage(attr.Value).unsafe = true
 			}
 		}
 		s.stack = append(s.stack, token.Name.Local)
@@ -230,7 +228,7 @@ func (s *editorReferenceScan) accept(token xml.Token) bool {
 				s.objects[id] = true
 			}
 		}
-		if (token.Name.Local == "Substitution" || token.Name.Local == "Font") && strings.TrimSpace(s.text.String()) != "" {
+		if token.Name.Local == "Font" && strings.TrimSpace(s.text.String()) != "" {
 			s.refs.fontUsage(s.text.String()).unsafe = true
 		}
 		if usage := s.fonts[len(s.fonts)-1]; usage != nil {
diff --git a/ofdgo_editor_resource_graph.go b/ofdgo_editor_resource_graph.go
new file mode 100644
index 0000000..8abc8cd
--- /dev/null
+++ b/ofdgo_editor_resource_graph.go
@@ -0,0 +1,103 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"encoding/xml"
+	"io"
+)
+
+// editorResourceElement 返回标准资源集合中的声明元素名称
+// 入参: group 资源集合名称
+// 返回: string 声明元素名称，未知集合返回空值
+func editorResourceElement(group string) string {
+	switch group {
+	case "Fonts":
+		return "Font"
+	case "MultiMedias":
+		return "MultiMedia"
+	case "ColorSpaces":
+		return "ColorSpace"
+	case "DrawParams":
+		return "DrawParam"
+	case "CompositeGraphicUnits":
+		return "CompositeGraphicUnit"
+	}
+	return ""
+}
+
+// editorResourceDefinitions 单次流式读取资源声明依赖，保留历史索引和非图像媒体
+// 入参: ctx 取消上下文, data 资源索引内容, name 文件路径, source 扫描结果, preserved 是否保留全部声明
+// 返回: *editorResourceRefs 含声明依赖的引用，无法安全分析时为空, error 解析或取消错误
+func editorResourceDefinitions(ctx context.Context, data []byte, name string, source *editorResourceRefs, preserved bool) (*editorResourceRefs, error) {
+	result := *source
+	result.ids, result.fonts = make(map[string]bool), nil
+	result.definitions = make(map[string]*editorResourceRefs)
+	outer := &editorResourceRefs{ids: make(map[string]bool), files: make(map[string]bool), fonts: make(map[string]*editorFontUsage)}
+	scan := editorReferenceScan{refs: outer, name: name, safe: true}
+	decoder := xml.NewDecoder(imageInput{ReadCloser: io.NopCloser(bytes.NewReader(data)), context: ctx})
+	for {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
+		token, err := decoder.Token()
+		if err == io.EOF {
+			if !scan.seen || len(scan.stack) != 0 {
+				return nil, nil
+			}
+			return &result, ctx.Err()
+		}
+		if err != nil {
+			return nil, err
+		}
+		depth := len(scan.stack)
+		if node, ok := token.(xml.StartElement); ok {
+			if depth == 0 && (scan.seen || node.Name.Local != "Res") || depth == 1 && editorResourceElement(node.Name.Local) == "" {
+				return nil, nil
+			}
+			if depth == 2 {
+				kind := editorResourceElement(scan.stack[1])
+				var id, mediaType string
+				for _, attr := range node.Attr {
+					if attr.Name.Space == "" {
+						switch attr.Name.Local {
+						case "ID":
+							id = editorResourceID(attr.Value)
+						case "Type":
+							mediaType = attr.Value
+						}
+					}
+				}
+				if node.Name.Local != kind || id == "" || result.definitions[id] != nil {
+					return nil, nil
+				}
+				refs := &editorResourceRefs{ids: make(map[string]bool), files: make(map[string]bool), fonts: make(map[string]*editorFontUsage)}
+				result.definitions[id] = refs
+				scan.refs = refs
+				if preserved || kind == "MultiMedia" && mediaType != "Image" {
+					result.ids[id] = true
+				}
+			}
+		}
+		if !scan.accept(token) {
+			return nil, ctx.Err()
+		}
+		if _, end := token.(xml.EndElement); end && depth == 3 {
+			scan.refs = outer
+		}
+	}
+}
diff --git a/ofdgo_editor_resource_scope.go b/ofdgo_editor_resource_scope.go
index c866886..85e0e7b 100644
--- a/ofdgo_editor_resource_scope.go
+++ b/ofdgo_editor_resource_scope.go
@@ -35,6 +35,7 @@ func resourceReferenceScopes(ctx context.Context, documents map[string]bool, res
 		scope := map[string]bool{entry: true}
 		queue := []string{entry}
 		refs := &editorResourceRefs{ids: make(map[string]bool), fonts: make(map[string]*editorFontUsage)}
+		definitions := make(map[string]*editorResourceRefs)
 		for position := 0; position < len(queue); position++ {
 			if err := ctx.Err(); err != nil {
 				return nil, err
@@ -47,6 +48,12 @@ func resourceReferenceScopes(ctx context.Context, documents map[string]bool, res
 			used[name] = true
 			if selected == "" || selected == entry {
 				refs.merge(&editorResourceRefs{ids: source.ids, fonts: source.fonts})
+				for id, definition := range source.definitions {
+					if definitions[id] != nil {
+						return nil, nil
+					}
+					definitions[id] = definition
+				}
 			}
 			for target, role := range source.roles {
 				if role&packageReferenceXML != 0 && !scope[target] {
@@ -58,6 +65,23 @@ func resourceReferenceScopes(ctx context.Context, documents map[string]bool, res
 		if selected != "" && selected != entry {
 			continue
 		}
+		var pending []string
+		if len(definitions) != 0 {
+			pending = slices.Collect(maps.Keys(refs.ids))
+		}
+		for position := 0; position < len(pending); position++ {
+			if err := ctx.Err(); err != nil {
+				return nil, err
+			}
+			if definition := definitions[pending[position]]; definition != nil {
+				for id := range definition.ids {
+					if !refs.ids[id] {
+						pending = append(pending, id)
+					}
+				}
+				refs.merge(&editorResourceRefs{ids: definition.ids, fonts: definition.fonts})
+			}
+		}
 		for name := range scope {
 			if err := ctx.Err(); err != nil {
 				return nil, err
diff --git a/ofdgo_editor_resources.go b/ofdgo_editor_resources.go
index 10c6169..1b78a6a 100644
--- a/ofdgo_editor_resources.go
+++ b/ofdgo_editor_resources.go
@@ -35,14 +35,15 @@ const (
 	editorReferenceFile
 )
 
-// editorResourceRefs 保存全包资源标识和直接文件引用，不以页面是否加载判断资源存活
+// editorResourceRefs 记录资源标识、文件引用及声明依赖，不以页面是否加载判断资源存活
 type editorResourceRefs struct {
-	ids       map[string]bool
-	files     map[string]bool
-	fonts     map[string]*editorFontUsage
-	versions  map[string]bool
-	documents map[string]bool
-	roles     map[string]int
+	ids         map[string]bool
+	files       map[string]bool
+	fonts       map[string]*editorFontUsage
+	versions    map[string]bool
+	documents   map[string]bool
+	roles       map[string]int
+	definitions map[string]*editorResourceRefs
 }
 
 // compactSourceReferences 合并自产页面引用，版本文档按当前版本范围扫描
@@ -51,7 +52,15 @@ type editorResourceRefs struct {
 // 入参: parts 输出条目, progress 保存进度回调, generated 自产页面及引用
 // 返回: map[string]bool 可移除条目, error 读取或取消错误
 func (e *Editor) compactSourceReferences(parts map[string][]byte, progress editorProgress, generated map[string]editorGeneratedReferences) (map[string]bool, error) {
-	multiple := e.source.reader.DocumentCount() > 1
+	return e.compactResourceReferences(parts, e.source.reader, progress, generated, e.source.retiredAttachments)
+}
+
+// compactResourceReferences 按文档引用清理输出副本，兼顾新建文档与已有包
+// 入参: parts 输出条目, reader 原包，可为空, progress 保存进度回调, generated 自产引用, retired 待移除附件
+// 返回: map[string]bool 可移除条目, error 读取或取消错误
+func (e *Editor) compactResourceReferences(parts map[string][]byte, reader *Reader, progress editorProgress, generated map[string]editorGeneratedReferences, retired map[string]bool) (map[string]bool, error) {
+	multiple := reader != nil && reader.DocumentCount() > 1
+	prune := e.output != nil && e.output.options.Mode != CompressionUnchanged
 	if e.output != nil && e.output.protected || multiple && (e.output == nil || e.output.options.Mode == CompressionUnchanged) {
 		return nil, nil
 	}
@@ -59,7 +68,6 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 		return nil, nil
 	}
 	generated = e.generatedVectorReferences(parts, generated)
-	reader := e.source.reader
 	ctx := context.Background()
 	if e.output != nil {
 		ctx = e.output.ctx
@@ -69,11 +77,13 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 		return nil, err
 	}
 	names := make(map[string]bool)
-	for name := range reader.fileIndex {
-		names[name] = true
-	}
-	for name := range reader.files {
-		names[name] = true
+	if reader != nil {
+		for name := range reader.fileIndex {
+			names[name] = true
+		}
+		for name := range reader.files {
+			names[name] = true
+		}
 	}
 	for name := range parts {
 		names[name] = true
@@ -85,7 +95,7 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 			}
 		}
 	}
-	refs, resources, references, safe, err := scanEditorResources(ctx, reader, parts, generated, names, progress, multiple)
+	refs, resources, references, safe, err := scanEditorResources(ctx, reader, parts, generated, names, progress, multiple || prune)
 	if err != nil {
 		return nil, err
 	}
@@ -93,9 +103,25 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 		return nil, nil
 	}
 	var scoped map[string]*editorResourceRefs
-	if multiple {
+	if prune {
+		for _, name := range resources {
+			data, ok := parts[name]
+			if !ok {
+				data, err = reader.readFile(name)
+				if err != nil {
+					return nil, err
+				}
+			}
+			indexed, err := editorResourceDefinitions(ctx, data, name, references[name], refs.versions[cleanPackagePath(name)])
+			if err != nil || indexed == nil {
+				return nil, err
+			}
+			references[name] = indexed
+		}
+	}
+	if multiple || prune {
 		selected := ""
-		if reader.OFD.DocBody[reader.documentIndex].versioned {
+		if reader != nil && reader.OFD.DocBody[reader.documentIndex].versioned {
 			selected, err = reader.DocumentRoot()
 			if err != nil {
 				return nil, err
@@ -107,11 +133,36 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 		}
 	}
 	ordered := slices.Sorted(maps.Keys(names))
-	removed := maps.Clone(e.source.retiredAttachments)
+	removed := maps.Clone(retired)
 	if removed == nil {
 		removed = make(map[string]bool)
 	}
-	usedFiles := maps.Clone(refs.files)
+	var usedFiles map[string]bool
+	var protectedProfiles map[string]bool
+	var profiles map[string][]editorProfileReference
+	if prune {
+		usedFiles = make(map[string]bool)
+		protectedProfiles = make(map[string]bool)
+		profiles = make(map[string][]editorProfileReference)
+		for name, source := range references {
+			if source.definitions == nil {
+				maps.Copy(usedFiles, source.files)
+				maps.Copy(protectedProfiles, source.files)
+				continue
+			}
+			usage := scoped[name]
+			for id, definition := range source.definitions {
+				if refs.versions[cleanPackagePath(name)] || usage == nil || usage.ids[id] {
+					maps.Copy(usedFiles, definition.files)
+					if refs.versions[cleanPackagePath(name)] || usage == nil {
+						maps.Copy(protectedProfiles, definition.files)
+					}
+				}
+			}
+		}
+	} else {
+		usedFiles = maps.Clone(refs.files)
+	}
 	updates := make(map[string][]byte)
 	fontFiles := make(map[string]*editorFontUsage)
 	for i, name := range resources {
@@ -131,7 +182,7 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 			return nil, nil
 		}
 		usageRefs := refs
-		if multiple {
+		if multiple || prune {
 			usageRefs = scoped[name]
 		}
 		preserved := refs.versions[cleanPackagePath(name)] || usageRefs == nil
@@ -145,35 +196,52 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 				} else if group.name.Local == "MultiMedias" && node.name.Local == "MultiMedia" {
 					field = "MediaFile"
 				}
-				if field == "" {
+				if field == "" && (!prune || node.name.Local != editorResourceElement(group.name.Local)) {
 					continue
 				}
 				id := editorResourceID(node.attr("ID"))
 				keep := preserved || id == "" || usageRefs.ids[id] || field == "MediaFile" && node.attr("Type") != "Image"
+				profile := prune && group.name.Local == "ColorSpaces" && node.name.Local == "ColorSpace"
+				if prune && keep && !profile {
+					maps.Copy(protectedProfiles, references[name].definitions[id].files)
+				}
+				var value string
 				if file := node.child(field); file != nil {
-					var value string
 					if err := xml.Unmarshal(data[file.start:file.end], &value); err != nil {
 						return nil, nil
 					}
-					if location := editorResourceLocation(name, root.attr("BaseLoc"), value); location != "" {
-						alternate := cleanPackagePath(resolveResourcePath(name, root.attr("BaseLoc"), value))
-						if field == "FontFile" && !multiple {
-							usage := usageRefs.fonts[id]
-							charset := node.attr("Charset")
-							unsafe := preserved || id == "" || charset != "" && !strings.EqualFold(charset, "unicode") || location != alternate
-							for _, file := range []string{location, alternate} {
-								if fontFiles[file] == nil {
-									fontFiles[file] = newEditorFontUsage()
-								}
-								fontFiles[file].merge(usage, unsafe || keep && usage == nil)
+				} else if profile {
+					value = node.attr("Profile")
+				}
+				if location := editorResourceLocation(name, root.attr("BaseLoc"), value); location != "" {
+					alternate := cleanPackagePath(resolveResourcePath(name, root.attr("BaseLoc"), value))
+					if field == "FontFile" && !multiple && keep {
+						var usage *editorFontUsage
+						if usageRefs != nil {
+							usage = usageRefs.fonts[id]
+						}
+						charset := node.attr("Charset")
+						unsafe := preserved || id == "" || charset != "" && !strings.EqualFold(charset, "unicode") || location != alternate
+						for _, file := range []string{location, alternate} {
+							if fontFiles[file] == nil {
+								fontFiles[file] = newEditorFontUsage()
+							}
+							fontFiles[file].merge(usage, unsafe || keep && usage == nil)
+						}
+					}
+					if keep {
+						usedFiles[location] = true
+						usedFiles[alternate] = true
+						if prune {
+							if profile && !preserved && location == alternate {
+								profiles[location] = append(profiles[location], editorProfileReference{index: name, id: id})
+							} else {
+								protectedProfiles[location] = true
+								protectedProfiles[alternate] = true
 							}
 						}
-						if keep {
-							usedFiles[location] = true
-							usedFiles[alternate] = true
-						} else {
-							removed[location] = true
-						}
+					} else {
+						removed[location] = true
 					}
 				}
 				if !keep {
@@ -226,6 +294,9 @@ func (e *Editor) compactSourceReferences(parts map[string][]byte, progress edito
 			delete(parts, name)
 		}
 	}
+	if err := compactResourceProfiles(ctx, reader, parts, profiles, protectedProfiles, removed); err != nil {
+		return nil, err
+	}
 	return removed, nil
 }
 
diff --git a/ofdgo_editor_versions.go b/ofdgo_editor_versions.go
index c54fa2d..ac55781 100644
--- a/ofdgo_editor_versions.go
+++ b/ofdgo_editor_versions.go
@@ -78,14 +78,25 @@ func (e *Editor) SelectVersion(id string) error {
 // 入参: name 版本名称，可为空
 // 返回: string 新版本标识, error 创建错误
 func (e *Editor) AddVersion(name string) (string, error) {
+	return e.AddVersionContext(context.Background(), name)
+}
+
+// AddVersionContext 以当前内容创建并选中默认版本，取消或失败时不改变文档和撤销历史
+// 初始共享文件，后续修改按需隔离，不自动写入创建日期
+// 入参: ctx 取消上下文, name 版本名称，可为空
+// 返回: string 新版本标识, error 创建或取消错误
+func (e *Editor) AddVersionContext(ctx context.Context, name string) (string, error) {
 	var id string
-	err := e.editDocuments(func(reader *Reader, data []byte, root *editorXML) (int, error) {
-		versions, err := reader.documentVersionEntries()
+	err := e.editDocumentsContext(ctx, func(reader *Reader, data []byte, root *editorXML) (int, error) {
+		versions, err := reader.documentVersionEntriesContext(ctx)
 		if err != nil {
 			return 0, err
 		}
 		index := int32(0)
 		for _, version := range versions {
+			if err := ctx.Err(); err != nil {
+				return 0, err
+			}
 			index = max(index, version.Index)
 		}
 		if index == math.MaxInt32 {
@@ -93,15 +104,26 @@ func (e *Editor) AddVersion(name string) (string, error) {
 		}
 		index++
 		ids := make(map[string]bool)
-		var collect func(*editorXML)
-		collect = func(node *editorXML) {
+		var collect func(*editorXML) error
+		collect = func(node *editorXML) error {
+			if err := ctx.Err(); err != nil {
+				return err
+			}
 			ids[node.attr("ID")] = true
 			for _, child := range node.children {
-				collect(child)
+				if err := collect(child); err != nil {
+					return err
+				}
 			}
+			return nil
+		}
+		if err := collect(root); err != nil {
+			return 0, err
 		}
-		collect(root)
 		for serial := int64(index); ; serial++ {
+			if err := ctx.Err(); err != nil {
+				return 0, err
+			}
 			id = "v" + strconv.FormatInt(serial, 10)
 			if !ids[id] {
 				break
@@ -111,7 +133,7 @@ func (e *Editor) AddVersion(name string) (string, error) {
 		if err != nil {
 			return 0, err
 		}
-		files, err := reader.versionSnapshotFiles(context.Background())
+		files, err := reader.versionSnapshotFiles(ctx)
 		if err != nil {
 			return 0, err
 		}
@@ -130,6 +152,10 @@ func (e *Editor) AddVersion(name string) (string, error) {
 			x.root("DocVersion", attrs)
 			x.start("FileList", nil)
 			for serial, file := range slices.Sorted(maps.Keys(files)) {
+				if err := ctx.Err(); err != nil {
+					x.err = err
+					return
+				}
 				x.start("File", ofdAttrs{{Name: xml.Name{Local: "ID"}, Value: "file" + strconv.Itoa(serial+1)}})
 				x.token(xml.CharData("/" + file))
 				x.end("File")
@@ -158,7 +184,7 @@ func (e *Editor) AddVersion(name string) (string, error) {
 		if err != nil {
 			return 0, err
 		}
-		root, err = parseEditorXML(data)
+		root, err = parseEditorXMLContext(ctx, data)
 		if err != nil {
 			return 0, err
 		}
@@ -262,7 +288,7 @@ func (e *Editor) RenameVersion(id, name string) error {
 			}
 		}
 		reader.files[location] = updated
-		if err := reader.retireVersionFiles(map[string]bool{version.Location: true}); err != nil {
+		if err := reader.retireVersionFiles(context.Background(), map[string]bool{version.Location: true}); err != nil {
 			return 0, err
 		}
 		return reader.documentIndex, nil
@@ -274,8 +300,16 @@ func (e *Editor) RenameVersion(id, name string) error {
 // 入参: id 版本标识，不能为空
 // 返回: error 删除错误
 func (e *Editor) DeleteVersion(id string) error {
-	return e.editDocuments(func(reader *Reader, data []byte, root *editorXML) (int, error) {
-		versions, err := reader.DocumentVersions()
+	return e.DeleteVersionContext(context.Background(), id)
+}
+
+// DeleteVersionContext 删除版本及其独占文件，取消或失败时保留当前状态和撤销历史
+// 文档主入口及其他版本仍在使用的文件不删除
+// 入参: ctx 取消上下文, id 版本标识，不能为空
+// 返回: error 删除或取消错误
+func (e *Editor) DeleteVersionContext(ctx context.Context, id string) error {
+	return e.editDocumentsContext(ctx, func(reader *Reader, data []byte, root *editorXML) (int, error) {
+		versions, err := reader.DocumentVersionsContext(ctx)
 		if err != nil {
 			return 0, err
 		}
@@ -289,6 +323,9 @@ func (e *Editor) DeleteVersion(id string) error {
 		}
 		list := root.childAt("DocBody", reader.documentIndex).child("Versions")
 		for _, node := range list.children {
+			if err := ctx.Err(); err != nil {
+				return 0, err
+			}
 			if node.matchesOFD("Version") && node.attr("ID") == id {
 				if len(versions) == 1 {
 					node = list
@@ -302,9 +339,12 @@ func (e *Editor) DeleteVersion(id string) error {
 		}
 		removed := map[string]bool{versions[index].Location: true}
 		for _, file := range versions[index].Files {
+			if err := ctx.Err(); err != nil {
+				return 0, err
+			}
 			removed[file.Location] = true
 		}
-		if err := reader.retireVersionFiles(removed); err != nil {
+		if err := reader.retireVersionFiles(ctx, removed); err != nil {
 			return 0, err
 		}
 		return reader.documentIndex, nil
@@ -341,6 +381,9 @@ func editorVersionDefault(data []byte, body *editorXML, id string) ([]byte, erro
 // 入参: ctx 取消上下文
 // 返回: map[string]bool 快照文件, error 引用或读取错误
 func (r *Reader) versionSnapshotFiles(ctx context.Context) (map[string]bool, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	entry, err := r.DocumentRoot()
 	if err != nil {
 		return nil, err
@@ -351,15 +394,18 @@ func (r *Reader) versionSnapshotFiles(ctx context.Context) (map[string]bool, err
 	}
 	if r.versionInfo != nil {
 		for _, file := range r.versionInfo.Files {
+			if err := ctx.Err(); err != nil {
+				return nil, err
+			}
 			files[file.Location] = true
 		}
 	}
 	body := r.OFD.DocBody[r.documentIndex]
-	data, err := r.readFile("OFD.xml")
+	data, err := r.readFileViewContext(ctx, "OFD.xml")
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -380,6 +426,9 @@ func (r *Reader) versionSnapshotFiles(ctx context.Context) (map[string]bool, err
 		maps.Copy(files, signed)
 	}
 	for name := range files {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		if !r.fileNames[name] {
 			return nil, fmt.Errorf("version file not found: %s", name)
 		}
@@ -388,24 +437,35 @@ func (r *Reader) versionSnapshotFiles(ctx context.Context) (map[string]bool, err
 }
 
 // retireVersionFiles 清理修改入口后不再被任何文档或版本引用的候选文件
-// 入参: candidates 待清理路径，OFD.xml始终保留
-// 返回: error 依赖读取错误
-func (r *Reader) retireVersionFiles(candidates map[string]bool) error {
+// 入参: ctx 取消上下文, candidates 待清理路径，OFD.xml始终保留
+// 返回: error 依赖读取或取消错误
+func (r *Reader) retireVersionFiles(ctx context.Context, candidates map[string]bool) error {
+	if err := ctx.Err(); err != nil {
+		return err
+	}
 	view := *r
 	if err := view.initRoot(); err != nil {
 		return err
 	}
-	for index := range view.DocumentCount() {
-		files, err := view.documentFiles(context.Background(), index)
-		if err != nil {
+	indexes := make([]int, view.DocumentCount())
+	for index := range indexes {
+		indexes[index] = index
+	}
+	files, err := view.documentFiles(ctx, indexes...)
+	if err != nil {
+		return err
+	}
+	for name := range files {
+		if err := ctx.Err(); err != nil {
 			return err
 		}
-		for name := range files {
-			delete(candidates, name)
-		}
+		delete(candidates, name)
 	}
 	delete(candidates, "OFD.xml")
 	for name := range candidates {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
 		delete(r.files, name)
 	}
 	if r.Zip != nil {
diff --git a/ofdgo_image.go b/ofdgo_image.go
index ba8c06d..f14305d 100644
--- a/ofdgo_image.go
+++ b/ofdgo_image.go
@@ -16,6 +16,7 @@ package ofdgo
 
 import (
 	"bytes"
+	"context"
 	"encoding/binary"
 	"fmt"
 	"image"
@@ -27,7 +28,7 @@ import (
 	_ "github.com/xiaoqidun/jbig2"
 	_ "golang.org/x/image/bmp"
 	"golang.org/x/image/draw"
-	_ "golang.org/x/image/tiff"
+	"golang.org/x/image/tiff"
 )
 
 // transparentEdgeImage 共享只读像素并按需计算透明边缘和采样边距
@@ -147,6 +148,14 @@ func decodeImageData(data []byte) (image.Image, string, error) {
 		}
 		return img, img.format, nil
 	}
+	if isTIFFData(data) {
+		layout, err := readTIFFLayout(bytes.NewReader(data))
+		if err != nil {
+			return nil, "tiff", err
+		}
+		img, err := decodeTIFFImage(nil, data, layout)
+		return img, "tiff", err
+	}
 	img, format, err := image.Decode(bytes.NewReader(data))
 	if err == nil {
 		return img, normalizeSealType(format), nil
@@ -175,6 +184,27 @@ func isPNGData(data []byte) bool {
 // 入参: data 图片数据
 // 返回: image.Config 图片尺寸, string 图片格式, error 错误信息
 func decodeImageConfigData(data []byte) (image.Config, string, error) {
+	if isTIFFData(data) {
+		layout, err := readTIFFLayout(bytes.NewReader(data))
+		if err != nil {
+			return image.Config{}, "tiff", err
+		}
+		if layout.planar && layout.samples > 1 || layout.samples == 1 && (layout.depth == 2 || layout.depth == 4) {
+			img, _, err := tiffPlaneDirectories(nil, data, layout)
+			if err != nil {
+				return image.Config{}, "tiff", err
+			}
+			size := layout.size()
+			return image.Config{ColorModel: img.ColorModel(), Width: size.X, Height: size.Y}, "tiff", nil
+		}
+		cfg, err := tiff.DecodeConfig(newTIFFDirectoryInput(data, layout.offset))
+		if err != nil {
+			return image.Config{}, "tiff", err
+		}
+		size := layout.size()
+		cfg.Width, cfg.Height = size.X, size.Y
+		return cfg, "tiff", nil
+	}
 	cfg, format, err := image.DecodeConfig(bytes.NewReader(data))
 	if err == nil {
 		return cfg, normalizeSealType(format), nil
@@ -193,11 +223,11 @@ func decodeBMPConfig(data []byte) (image.Config, error) {
 	if err != nil {
 		return image.Config{}, err
 	}
-	if offset <= 0 || width <= 0 || height == 0 || compression != 0 {
+	if offset <= 0 || width <= 0 || height == 0 {
 		return image.Config{}, fmt.Errorf("unsupported bmp")
 	}
-	if bpp != 16 && bpp != 24 && bpp != 32 {
-		return image.Config{}, fmt.Errorf("unsupported bmp")
+	if _, err := bmpColorMasks(data, offset, bpp, compression); err != nil {
+		return image.Config{}, err
 	}
 	if height < 0 {
 		height = -height
@@ -209,15 +239,26 @@ func decodeBMPConfig(data []byte) (image.Config, error) {
 // 入参: data 图片数据
 // 返回: image.Image 图片对象, error 错误信息
 func decodeBMPImage(data []byte) (image.Image, error) {
+	return decodeBMPImageContext(context.Background(), data)
+}
+
+// decodeBMPImageContext 解码BMP真彩色像素，逐行检查取消
+// 入参: ctx 取消上下文, data 图片数据
+// 返回: image.Image 图片对象, error 格式或取消错误
+func decodeBMPImageContext(ctx context.Context, data []byte) (image.Image, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	offset, width, height, bpp, compression, err := parseBMPHeader(data)
 	if err != nil {
 		return nil, err
 	}
-	if compression != 0 {
-		return nil, fmt.Errorf("unsupported bmp compression")
+	if offset > len(data) {
+		return nil, fmt.Errorf("invalid bmp pixel offset")
 	}
-	if bpp != 16 && bpp != 24 && bpp != 32 {
-		return nil, fmt.Errorf("unsupported bmp depth")
+	masks, err := bmpColorMasks(data, offset, bpp, compression)
+	if err != nil {
+		return nil, err
 	}
 	topDown := height < 0
 	if topDown {
@@ -230,9 +271,15 @@ func decodeBMPImage(data []byte) (image.Image, error) {
 	if stride > int64(len(data)-offset)/int64(height) {
 		return nil, fmt.Errorf("truncated bmp")
 	}
+	if width > int(^uint(0)>>1)/4/height {
+		return nil, fmt.Errorf("bmp dimensions exceed image capacity")
+	}
 	rowStride := int(stride)
 	img := image.NewNRGBA(image.Rect(0, 0, width, height))
 	for y := 0; y < height; y++ {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		srcY := y
 		if !topDown {
 			srcY = height - 1 - y
@@ -240,20 +287,20 @@ func decodeBMPImage(data []byte) (image.Image, error) {
 		row := data[offset+srcY*rowStride:]
 		for x := 0; x < width; x++ {
 			switch bpp {
-			case 16:
-				v := binary.LittleEndian.Uint16(row[x*2:])
-				img.SetNRGBA(x, y, color.NRGBA{
-					R: uint8(((v >> 10) & 0x1F) * 255 / 31),
-					G: uint8(((v >> 5) & 0x1F) * 255 / 31),
-					B: uint8((v & 0x1F) * 255 / 31),
-					A: 255,
-				})
+			case 16, 32:
+				var value uint32
+				if bpp == 16 {
+					value = uint32(binary.LittleEndian.Uint16(row[x*2:]))
+				} else {
+					value = binary.LittleEndian.Uint32(row[x*4:])
+				}
+				p := y*img.Stride + x*4
+				for i, mask := range masks {
+					img.Pix[p+i] = mask.sample(value)
+				}
 			case 24:
 				p := x * 3
 				img.SetNRGBA(x, y, color.NRGBA{R: row[p+2], G: row[p+1], B: row[p], A: 255})
-			case 32:
-				p := x * 4
-				img.SetNRGBA(x, y, color.NRGBA{R: row[p+2], G: row[p+1], B: row[p], A: 255})
 			}
 		}
 	}
@@ -272,7 +319,7 @@ func parseBMPHeader(data []byte) (int, int, int, uint16, uint32, error) {
 	if dibSize < 40 {
 		return 0, 0, 0, 0, 0, fmt.Errorf("unsupported bmp")
 	}
-	if uint64(dibSize)+14 > uint64(len(data)) || offset < 0 || uint64(offset) < uint64(dibSize)+14 || offset > len(data) {
+	if uint64(dibSize)+14 > uint64(len(data)) || offset < 0 || uint64(offset) < uint64(dibSize)+14 {
 		return 0, 0, 0, 0, 0, fmt.Errorf("invalid bmp pixel offset")
 	}
 	width := int(int32(binary.LittleEndian.Uint32(data[18:22])))
@@ -450,6 +497,9 @@ func imageNRGBA64At(img image.Image, x, y int) color.NRGBA64 {
 	switch src := img.(type) {
 	case interface{ NRGBA64At(int, int) color.NRGBA64 }:
 		return src.NRGBA64At(x, y)
+	case *image.NRGBA:
+		c := src.NRGBAAt(x, y)
+		return color.NRGBA64{R: uint16(c.R) * 257, G: uint16(c.G) * 257, B: uint16(c.B) * 257, A: uint16(c.A) * 257}
 	case *image.Gray16:
 		c := src.Gray16At(x, y)
 		return color.NRGBA64{R: c.Y, G: c.Y, B: c.Y, A: 0xffff}
diff --git a/ofdgo_image_bmp.go b/ofdgo_image_bmp.go
new file mode 100644
index 0000000..a0d345c
--- /dev/null
+++ b/ofdgo_image_bmp.go
@@ -0,0 +1,125 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bufio"
+	"encoding/binary"
+	"fmt"
+	"image"
+	"io"
+	"math/bits"
+)
+
+// decodeImageConfig 读取图片尺寸，BMP和TIFF只读取布局信息，不解码像素
+// 入参: input 图片输入流
+// 返回: image.Config 图片配置, string 图片格式, error 头部错误
+func decodeImageConfig(input io.Reader) (image.Config, string, error) {
+	buffer := bufio.NewReader(input)
+	prefix, _ := buffer.Peek(34)
+	if isTIFFData(prefix) {
+		layout, err := readTIFFLayout(buffer)
+		if err != nil {
+			return image.Config{}, "tiff", err
+		}
+		size := layout.size()
+		return image.Config{Width: size.X, Height: size.Y}, "tiff", nil
+	}
+	if len(prefix) == 34 && string(prefix[:2]) == "BM" {
+		header := binary.LittleEndian.Uint32(prefix[14:18])
+		bpp := binary.LittleEndian.Uint16(prefix[28:30])
+		compression := binary.LittleEndian.Uint32(prefix[30:34])
+		if (bpp == 16 || bpp == 32 && compression == 3) && (header == 40 || header == 52 || header == 56 || header == 108 || header == 124) {
+			size := int(header) + 14
+			if header == 40 && compression == 3 {
+				size += 12
+			}
+			data := make([]byte, size)
+			if _, err := io.ReadFull(buffer, data); err != nil {
+				return image.Config{}, "bmp", err
+			}
+			config, err := decodeBMPConfig(data)
+			return config, "bmp", err
+		}
+	}
+	return image.DecodeConfig(buffer)
+}
+
+// bmpColorMask 保存一个BMP颜色分量的位域位置和取值范围
+type bmpColorMask struct {
+	mask  uint32
+	shift uint
+	max   uint32
+}
+
+// sample 将位域采样值映射到八位通道，缺省透明度为不透明
+// 入参: value 编码像素
+// 返回: uint8 通道值
+func (m bmpColorMask) sample(value uint32) uint8 {
+	if m.mask == 0 {
+		return 255
+	}
+	return uint8(uint64((value&m.mask)>>m.shift) * 255 / uint64(m.max))
+}
+
+// bmpColorMasks 读取BMP真彩色通道布局，校验位域连续性、重叠及像素范围
+// 入参: data 已校验头部的BMP数据, offset 像素偏移, bpp 像素位数, compression 编码方式
+// 返回: [4]bmpColorMask 红绿蓝和透明度通道, error 格式错误
+func bmpColorMasks(data []byte, offset int, bpp uint16, compression uint32) ([4]bmpColorMask, error) {
+	var result [4]bmpColorMask
+	if bpp != 16 && bpp != 24 && bpp != 32 {
+		return result, fmt.Errorf("unsupported bmp depth")
+	}
+	masks := [4]uint32{0xff0000, 0xff00, 0xff}
+	if bpp == 16 {
+		masks = [4]uint32{0x7c00, 0x3e0, 0x1f}
+	}
+	switch compression {
+	case 0:
+	case 3:
+		if bpp == 24 {
+			return result, fmt.Errorf("invalid bmp bitfield depth")
+		}
+		header := binary.LittleEndian.Uint32(data[14:18])
+		if offset < 66 || len(data) < 66 || header != 40 && header < 52 {
+			return result, fmt.Errorf("truncated bmp color masks")
+		}
+		for i := range 3 {
+			masks[i] = binary.LittleEndian.Uint32(data[54+i*4:])
+		}
+		if header >= 56 {
+			masks[3] = binary.LittleEndian.Uint32(data[66:70])
+		}
+	default:
+		return result, fmt.Errorf("unsupported bmp compression")
+	}
+	var used uint32
+	for i, mask := range masks {
+		if mask == 0 {
+			if i < 3 {
+				return result, fmt.Errorf("empty bmp color mask")
+			}
+			continue
+		}
+		shift := uint(bits.TrailingZeros32(mask))
+		maximum := mask >> shift
+		if mask&used != 0 || maximum&(maximum+1) != 0 || uint64(mask)>>bpp != 0 {
+			return result, fmt.Errorf("invalid bmp color mask")
+		}
+		result[i] = bmpColorMask{mask: mask, shift: shift, max: maximum}
+		used |= mask
+	}
+	return result, nil
+}
diff --git a/ofdgo_image_metadata.go b/ofdgo_image_metadata.go
index 9153711..ac3b55c 100644
--- a/ofdgo_image_metadata.go
+++ b/ofdgo_image_metadata.go
@@ -27,12 +27,28 @@ type imagePageInfo struct {
 	xdpi, ydpi  float64
 }
 
-// imagePageMetadata 读取PNG物理尺寸与JPEG的JFIF、EXIF信息，EXIF密度优先于JFIF
+// imagePageMetadata 读取图片方向与密度，TIFF方向由解码器处理，JPEG的EXIF密度优先于JFIF
 // 入参: data 图片数据, format 图像格式, dpi 缺少物理密度时使用的DPI
 // 返回: imagePageInfo 方向与横纵DPI, error 元数据错误
 func imagePageMetadata(data []byte, format string, dpi float64) (imagePageInfo, error) {
 	info := imagePageInfo{orientation: 1, xdpi: dpi, ydpi: dpi}
 	var exif []byte
+	var directory uint32
+	if format == "tiff" {
+		layout, err := readTIFFLayout(bytes.NewReader(data))
+		if err != nil {
+			return info, err
+		}
+		directory = layout.offset
+		exif = data
+	}
+	if format == "bmp" && len(data) >= 54 && binary.LittleEndian.Uint32(data[14:18]) >= 40 {
+		x := int32(binary.LittleEndian.Uint32(data[38:42]))
+		y := int32(binary.LittleEndian.Uint32(data[42:46]))
+		if x > 0 && y > 0 {
+			info.xdpi, info.ydpi = float64(x)*0.0254, float64(y)*0.0254
+		}
+	}
 	if format == "png" {
 		for offset := 8; offset+12 <= len(data); {
 			n := uint64(binary.BigEndian.Uint32(data[offset:]))
@@ -109,22 +125,28 @@ func imagePageMetadata(data []byte, format string, dpi float64) (imagePageInfo,
 		}
 	}
 	if len(exif) != 0 {
-		orientation, x, y, err := imageExifMetadata(exif)
+		orientation, x, y, err := imageExifMetadata(exif, dpi, directory)
 		if err != nil {
 			return info, err
 		}
 		info.orientation = orientation
-		if format == "jpeg" && x > 0 && y > 0 {
+		if (format == "jpeg" || format == "tiff") && x > 0 && y > 0 {
 			info.xdpi, info.ydpi = x, y
 		}
+		if format == "tiff" {
+			if orientation >= 5 {
+				info.xdpi, info.ydpi = info.ydpi, info.xdpi
+			}
+			info.orientation = 1
+		}
 	}
 	return info, nil
 }
 
 // imageExifMetadata 只读取主图IFD中的方向和密度，不跟随缩略图与其他目录
-// 入参: data EXIF中的TIFF数据
+// 入参: data TIFF或EXIF数据, dpi 无绝对密度单位时的横向DPI，零值不换算, directory 目录偏移，零值使用首目录
 // 返回: int EXIF方向值，未声明时为1, float64 横向DPI, float64 纵向DPI, error 元数据错误
-func imageExifMetadata(data []byte) (int, float64, float64, error) {
+func imageExifMetadata(data []byte, dpi float64, directory uint32) (int, float64, float64, error) {
 	invalid := fmt.Errorf("invalid EXIF image metadata")
 	if len(data) < 8 {
 		return 0, 0, 0, invalid
@@ -142,6 +164,9 @@ func imageExifMetadata(data []byte) (int, float64, float64, error) {
 		return 0, 0, 0, invalid
 	}
 	offset := uint64(order.Uint32(data[4:]))
+	if directory != 0 {
+		offset = uint64(directory)
+	}
 	if offset < 8 || offset+2 > uint64(len(data)) {
 		return 0, 0, 0, invalid
 	}
@@ -185,6 +210,8 @@ func imageExifMetadata(data []byte) (int, float64, float64, error) {
 	}
 	if unit == 3 {
 		x, y = x*2.54, y*2.54
+	} else if unit == 1 && dpi > 0 && x > 0 && y > 0 {
+		x, y = dpi, dpi*y/x
 	} else if unit != 2 {
 		x, y = 0, 0
 	}
diff --git a/ofdgo_image_tiff.go b/ofdgo_image_tiff.go
new file mode 100644
index 0000000..a26b492
--- /dev/null
+++ b/ofdgo_image_tiff.go
@@ -0,0 +1,354 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"encoding/binary"
+	"errors"
+	"fmt"
+	"image"
+	"image/color"
+	"io"
+)
+
+// tiffImageLayout 保存单页TIFF的尺寸、方向、采样及压缩参数
+type tiffImageLayout struct {
+	width, height int
+	orientation   int
+	offset, next  uint32
+	reduced, mask bool
+	planar        bool
+	samples       uint32
+	depth         uint32
+	fillOrder     uint32
+	compression   uint32
+	t4, t6        uint32
+}
+
+// errTIFFRandomAccess 表示目录链需要回读，仅流式输入需重新打开资源
+var errTIFFRandomAccess = errors.New("TIFF directory requires random access")
+
+// tiffDirectoryInput 只替换解码视图中的首目录偏移，不复制或改写原始资源
+type tiffDirectoryInput struct {
+	reader    *bytes.Reader
+	directory [4]byte
+	ctx       context.Context
+}
+
+// orientedTIFFImage 按方向映射像素坐标，复用原图数据与颜色精度
+type orientedTIFFImage struct {
+	source image.Image
+	tiffImageLayout
+}
+
+// isTIFFData 识别标准TIFF字节序与版本标识
+// 入参: data 文件头或完整编码
+// 返回: bool 是否为TIFF 6.0编码
+func isTIFFData(data []byte) bool {
+	return len(data) >= 4 && (string(data[:4]) == "II\x2a\x00" || string(data[:4]) == "MM\x00\x2a")
+}
+
+// readTIFFLayout 选择单页TIFF的主图，允许缩略图目录并拒绝循环和多个主图
+// 入参: input 从文件头开始的输入流
+// 返回: tiffImageLayout 存储尺寸与方向, error 文件结构或方向错误
+func readTIFFLayout(input io.Reader) (tiffImageLayout, error) {
+	var result tiffImageLayout
+	var header [8]byte
+	if _, err := io.ReadFull(input, header[:]); err != nil {
+		return result, err
+	}
+	if !isTIFFData(header[:]) {
+		return result, fmt.Errorf("invalid TIFF header")
+	}
+	var order binary.ByteOrder = binary.LittleEndian
+	if header[0] == 'M' {
+		order = binary.BigEndian
+	}
+	offset := int64(order.Uint32(header[4:]))
+	position := int64(8)
+	var seen map[uint32]bool
+	for offset != 0 {
+		if seen != nil && seen[uint32(offset)] {
+			return result, fmt.Errorf("cyclic TIFF directory chain")
+		}
+		layout, end, err := readTIFFDirectory(input, order, offset, position)
+		if err != nil {
+			return result, err
+		}
+		if layout.mask {
+			return result, fmt.Errorf("TIFF transparency mask directories are not supported")
+		}
+		if !layout.reduced || layout.next == 0 && seen == nil {
+			if result.offset != 0 {
+				return result, fmt.Errorf("multiple TIFF pages are not supported in OFD resources")
+			}
+			result = layout
+		}
+		if layout.next != 0 {
+			if seen == nil {
+				seen = make(map[uint32]bool)
+			}
+			seen[uint32(offset)] = true
+		}
+		offset, position = int64(layout.next), end
+	}
+	if result.offset == 0 {
+		return result, fmt.Errorf("missing TIFF main image")
+	}
+	return result, nil
+}
+
+// readTIFFDirectory 读取目录中的图像布局和子文件类型，不读取像素
+// 入参: input 输入流, order 字节序, offset 目录偏移, position 当前流位置
+// 返回: tiffImageLayout 目录布局, int64 读取后的流位置, error 结构或读取错误
+func readTIFFDirectory(input io.Reader, order binary.ByteOrder, offset, position int64) (tiffImageLayout, int64, error) {
+	var result tiffImageLayout
+	var header [4]byte
+	if offset < 8 {
+		return result, position, fmt.Errorf("invalid TIFF directory offset")
+	}
+	if seeker, ok := input.(io.Seeker); ok {
+		if _, err := seeker.Seek(offset, io.SeekStart); err != nil {
+			return result, position, err
+		}
+	} else {
+		if offset < position {
+			return result, position, errTIFFRandomAccess
+		}
+		if _, err := io.CopyN(io.Discard, input, offset-position); err != nil {
+			return result, position, err
+		}
+	}
+	if _, err := io.ReadFull(input, header[:2]); err != nil {
+		return result, position, err
+	}
+	count := int(order.Uint16(header[:2]))
+	result.orientation, result.offset, result.samples, result.fillOrder = 1, uint32(offset), 1, 1
+	result.depth = 1
+	result.compression = 1
+	var entry [12]byte
+	var previous uint16
+	var subfile, legacy uint32
+	var hasSubfile bool
+	for i := range count {
+		if _, err := io.ReadFull(input, entry[:]); err != nil {
+			return result, position, err
+		}
+		tag, kind, number := order.Uint16(entry[:2]), order.Uint16(entry[2:4]), order.Uint32(entry[4:8])
+		if i > 0 && tag <= previous {
+			return result, position, fmt.Errorf("invalid TIFF directory order")
+		}
+		previous = tag
+		if tag == 258 && kind == 3 && number == 1 {
+			result.depth = uint32(order.Uint16(entry[8:10]))
+		}
+		if tag != 254 && tag != 255 && tag != 256 && tag != 257 && tag != 259 && tag != 262 && tag != 266 && tag != 274 && tag != 277 && tag != 284 && tag != 292 && tag != 293 {
+			continue
+		}
+		if number != 1 || kind != 3 && kind != 4 || (tag == 255 || tag == 259 || tag == 262 || tag == 266 || tag == 274 || tag == 277 || tag == 284) && kind != 3 || (tag == 254 || tag == 292 || tag == 293) && kind != 4 {
+			return result, position, fmt.Errorf("invalid TIFF image layout")
+		}
+		value := uint64(order.Uint32(entry[8:12]))
+		if kind == 3 {
+			value = uint64(order.Uint16(entry[8:10]))
+		}
+		if (tag == 256 || tag == 257) && (value == 0 || value > uint64(int(^uint(0)>>1))) {
+			return result, position, fmt.Errorf("invalid TIFF dimensions")
+		}
+		switch tag {
+		case 254:
+			subfile, hasSubfile = uint32(value), true
+		case 255:
+			legacy = uint32(value)
+		case 256:
+			result.width = int(value)
+		case 257:
+			result.height = int(value)
+		case 259:
+			result.compression = uint32(value)
+		case 262:
+			result.mask = value == 4
+		case 266:
+			if value != 1 && value != 2 {
+				return result, position, fmt.Errorf("invalid TIFF fill order")
+			}
+			result.fillOrder = uint32(value)
+		case 274:
+			if value < 1 || value > 8 {
+				return result, position, fmt.Errorf("invalid TIFF orientation")
+			}
+			result.orientation = int(value)
+		case 277:
+			if value == 0 {
+				return result, position, fmt.Errorf("invalid TIFF sample count")
+			}
+			result.samples = uint32(value)
+		case 284:
+			if value != 1 && value != 2 {
+				return result, position, fmt.Errorf("invalid TIFF planar configuration")
+			}
+			result.planar = value == 2
+		case 292:
+			result.t4 = uint32(value)
+		case 293:
+			result.t6 = uint32(value)
+		}
+	}
+	if _, err := io.ReadFull(input, header[:4]); err != nil {
+		return result, position, err
+	}
+	result.next = order.Uint32(header[:4])
+	if hasSubfile {
+		result.reduced, result.mask = subfile&1 != 0, result.mask || subfile&4 != 0
+	} else {
+		result.reduced = legacy == 2
+	}
+	if result.width == 0 || result.height == 0 {
+		return result, position, fmt.Errorf("missing TIFF dimensions")
+	}
+	return result, offset + 6 + int64(count)*12, nil
+}
+
+// newTIFFDirectoryInput 创建以指定目录作为主图的只读解码视图
+// 入参: data 已校验的TIFF编码, offset 主图目录偏移
+// 返回: *tiffDirectoryInput 随机访问输入
+func newTIFFDirectoryInput(data []byte, offset uint32) *tiffDirectoryInput {
+	input := &tiffDirectoryInput{reader: bytes.NewReader(data)}
+	var order binary.ByteOrder = binary.LittleEndian
+	if data[0] == 'M' {
+		order = binary.BigEndian
+	}
+	order.PutUint32(input.directory[:], offset)
+	return input
+}
+
+// patch 替换读取片段中与首目录指针重叠的部分
+// 入参: data 读取结果, offset 文件偏移
+func (r *tiffDirectoryInput) patch(data []byte, offset int64) {
+	for i := range r.directory {
+		at := int64(4+i) - offset
+		if at >= 0 && at < int64(len(data)) {
+			data[at] = r.directory[i]
+		}
+	}
+}
+
+// Read 读取图像编码并覆盖首目录指针
+// 入参: data 接收缓冲
+// 返回: int 读取字节数, error 读取错误
+func (r *tiffDirectoryInput) Read(data []byte) (int, error) {
+	if r.ctx != nil {
+		if err := r.ctx.Err(); err != nil {
+			return 0, err
+		}
+	}
+	offset := r.reader.Size() - int64(r.reader.Len())
+	n, err := r.reader.Read(data)
+	r.patch(data[:n], offset)
+	return n, err
+}
+
+// ReadAt 随机读取图像编码并覆盖首目录指针
+// 入参: data 接收缓冲, offset 文件偏移
+// 返回: int 读取字节数, error 读取错误
+func (r *tiffDirectoryInput) ReadAt(data []byte, offset int64) (int, error) {
+	if r.ctx != nil {
+		if err := r.ctx.Err(); err != nil {
+			return 0, err
+		}
+	}
+	n, err := r.reader.ReadAt(data, offset)
+	r.patch(data[:n], offset)
+	return n, err
+}
+
+// size 返回应用方向后的显示尺寸
+// 返回: image.Point 像素宽高
+func (l tiffImageLayout) size() image.Point {
+	if l.orientation >= 5 {
+		return image.Pt(l.height, l.width)
+	}
+	return image.Pt(l.width, l.height)
+}
+
+// Bounds 返回从零开始的显示区域
+// 返回: image.Rectangle 显示区域
+func (s *orientedTIFFImage) Bounds() image.Rectangle {
+	return image.Rectangle{Max: s.size()}
+}
+
+// ColorModel 返回原始颜色模型
+// 返回: color.Model 颜色模型
+func (s *orientedTIFFImage) ColorModel() color.Model { return s.source.ColorModel() }
+
+// sourcePoint 将显示坐标映射为存储坐标
+// 入参: x 横向显示坐标, y 纵向显示坐标
+// 返回: image.Point 原始坐标, bool 是否位于显示区域内
+func (s *orientedTIFFImage) sourcePoint(x, y int) (image.Point, bool) {
+	if !image.Pt(x, y).In(s.Bounds()) {
+		return image.Point{}, false
+	}
+	w, h := s.width, s.height
+	switch s.orientation {
+	case 2:
+		x = w - 1 - x
+	case 3:
+		x, y = w-1-x, h-1-y
+	case 4:
+		y = h - 1 - y
+	case 5:
+		x, y = y, x
+	case 6:
+		x, y = y, h-1-x
+	case 7:
+		x, y = w-1-y, h-1-x
+	case 8:
+		x, y = w-1-y, x
+	}
+	b := s.source.Bounds()
+	return image.Pt(b.Min.X+x, b.Min.Y+y), true
+}
+
+// At 返回原图颜色，不重新采样或量化
+// 入参: x 横向显示坐标, y 纵向显示坐标
+// 返回: color.Color 像素颜色，区域外透明
+func (s *orientedTIFFImage) At(x, y int) color.Color {
+	p, ok := s.sourcePoint(x, y)
+	if !ok {
+		return color.NRGBA64{}
+	}
+	return s.source.At(p.X, p.Y)
+}
+
+// NRGBA64At 读取非预乘十六位颜色，保留透明像素中的原始颜色
+// 入参: x 横向显示坐标, y 纵向显示坐标
+// 返回: color.NRGBA64 像素颜色
+func (s *orientedTIFFImage) NRGBA64At(x, y int) color.NRGBA64 {
+	p, ok := s.sourcePoint(x, y)
+	if !ok {
+		return color.NRGBA64{}
+	}
+	return imageNRGBA64At(s.source, p.X, p.Y)
+}
+
+// RGBA64At 返回预乘十六位像素，保留原始颜色精度
+// 入参: x 横向显示坐标, y 纵向显示坐标
+// 返回: color.RGBA64 像素颜色
+func (s *orientedTIFFImage) RGBA64At(x, y int) color.RGBA64 {
+	r, g, b, a := s.At(x, y).RGBA()
+	return color.RGBA64{R: uint16(r), G: uint16(g), B: uint16(b), A: uint16(a)}
+}
diff --git a/ofdgo_image_tiff_bits.go b/ofdgo_image_tiff_bits.go
new file mode 100644
index 0000000..1bea5ab
--- /dev/null
+++ b/ofdgo_image_tiff_bits.go
@@ -0,0 +1,189 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"cmp"
+	"context"
+	"encoding/binary"
+	"fmt"
+	"image"
+	"io"
+	"math/bits"
+	"slices"
+	"sort"
+
+	"golang.org/x/image/tiff"
+)
+
+// tiffBitOrderInput 仅在读取像素块时反转字节内位序，保留目录和原始数据
+type tiffBitOrderInput struct {
+	source   io.ReaderAt
+	ranges   [][2]int64
+	position int64
+}
+
+// tiffBitOrderRanges 收集低位优先的未压缩一位图像块，压缩传真位序由解码器处理
+// 入参: ctx 取消上下文，可为nil, data TIFF编码, layout 已校验的主图布局
+// 返回: [][2]int64 按偏移排序且不重叠的字节区间, error 字段或布局错误
+func tiffBitOrderRanges(ctx context.Context, data []byte, layout tiffImageLayout) ([][2]int64, error) {
+	if layout.fillOrder != 2 {
+		return nil, nil
+	}
+	var order binary.ByteOrder = binary.LittleEndian
+	if data[0] == 'M' {
+		order = binary.BigEndian
+	}
+	offset := int(layout.offset)
+	count := int(order.Uint16(data[offset:]))
+	entries := make(map[uint16][]byte, count)
+	for i := range count {
+		entry := data[offset+2+i*12 : offset+14+i*12]
+		entries[order.Uint16(entry)] = entry
+	}
+	compression := entries[259]
+	method := uint16(1)
+	if compression != nil {
+		if order.Uint16(compression[2:]) != 3 || order.Uint32(compression[4:]) != 1 {
+			return nil, fmt.Errorf("invalid TIFF compression field")
+		}
+		method = order.Uint16(compression[8:])
+	}
+	if method == 3 || method == 4 {
+		return nil, nil
+	}
+	if method != 1 {
+		return nil, fmt.Errorf("unsupported TIFF fill order for compression %d", method)
+	}
+	if entry := entries[258]; entry != nil {
+		if order.Uint16(entry[2:]) != 3 || order.Uint32(entry[4:]) != layout.samples {
+			return nil, fmt.Errorf("invalid TIFF sample depths")
+		}
+		for i := uint32(0); i < layout.samples; i++ {
+			depth, err := tiffPlaneValue(data, order, entry, i)
+			if err != nil || depth != 1 {
+				return nil, fmt.Errorf("unsupported TIFF fill order for sample depth")
+			}
+		}
+	}
+	locations, lengths := entries[273], entries[279]
+	value := func(tag uint16, fallback uint32) (uint32, error) {
+		entry := entries[tag]
+		if entry == nil {
+			return fallback, nil
+		}
+		if order.Uint32(entry[4:]) != 1 {
+			return 0, fmt.Errorf("invalid TIFF scalar field %d", tag)
+		}
+		return tiffPlaneValue(data, order, entry, 0)
+	}
+	rows, err := value(278, ^uint32(0))
+	if err != nil || rows == 0 {
+		return nil, fmt.Errorf("invalid TIFF rows per strip")
+	}
+	expected := (uint64(layout.height)-1)/uint64(rows) + 1
+	if entries[322] != nil {
+		locations, lengths = entries[324], entries[325]
+		width, err := value(322, 0)
+		if err != nil || width == 0 || width%16 != 0 {
+			return nil, fmt.Errorf("invalid TIFF tile width")
+		}
+		height, err := value(323, 0)
+		if err != nil || height == 0 || height%16 != 0 {
+			return nil, fmt.Errorf("invalid TIFF tile height")
+		}
+		expected = ((uint64(layout.width)-1)/uint64(width) + 1) * ((uint64(layout.height)-1)/uint64(height) + 1)
+		if locations != nil && order.Uint16(locations[2:]) != 4 {
+			return nil, fmt.Errorf("invalid TIFF tile offsets")
+		}
+	}
+	if layout.planar {
+		if expected > uint64(^uint32(0))/uint64(layout.samples) {
+			return nil, fmt.Errorf("too many TIFF image blocks")
+		}
+		expected *= uint64(layout.samples)
+	}
+	if locations == nil || lengths == nil {
+		return nil, fmt.Errorf("missing TIFF block arrays")
+	}
+	blocks := order.Uint32(locations[4:])
+	if blocks == 0 || uint64(blocks) != expected || order.Uint32(lengths[4:]) != blocks {
+		return nil, fmt.Errorf("invalid TIFF block count")
+	}
+	var ranges [][2]int64
+	for i := uint32(0); i < blocks; i++ {
+		if ctx != nil {
+			if err := ctx.Err(); err != nil {
+				return nil, err
+			}
+		}
+		at, err := tiffPlaneValue(data, order, locations, i)
+		if err != nil {
+			return nil, err
+		}
+		size, err := tiffPlaneValue(data, order, lengths, i)
+		if err != nil || size == 0 || uint64(at)+uint64(size) > uint64(len(data)) {
+			return nil, fmt.Errorf("truncated TIFF image block")
+		}
+		ranges = append(ranges, [2]int64{int64(at), int64(at) + int64(size)})
+	}
+	slices.SortFunc(ranges, func(a, b [2]int64) int { return cmp.Compare(a[0], b[0]) })
+	merged := ranges[:0]
+	for _, span := range ranges {
+		if len(merged) != 0 && span[0] <= merged[len(merged)-1][1] {
+			merged[len(merged)-1][1] = max(merged[len(merged)-1][1], span[1])
+		} else {
+			merged = append(merged, span)
+		}
+	}
+	return merged, nil
+}
+
+// decodeTIFFBitOrder 使用只读视图校正位序，不分配整幅编码的副本
+// 入参: input 原始解码视图, ranges 需要反转位序的像素区间
+// 返回: image.Image 解码图像, error 解码错误
+func decodeTIFFBitOrder(input interface {
+	io.Reader
+	io.ReaderAt
+}, ranges [][2]int64) (image.Image, error) {
+	if len(ranges) == 0 {
+		return tiff.Decode(input)
+	}
+	return tiff.Decode(&tiffBitOrderInput{source: input, ranges: ranges})
+}
+
+// Read 顺序读取并校正像素字节的位序
+// 入参: data 接收缓冲
+// 返回: int 读取字节数, error 读取错误
+func (r *tiffBitOrderInput) Read(data []byte) (int, error) {
+	n, err := r.ReadAt(data, r.position)
+	r.position += int64(n)
+	return n, err
+}
+
+// ReadAt 按文件偏移读取，只反转实际读取到的像素字节
+// 入参: data 接收缓冲, offset 文件偏移
+// 返回: int 读取字节数, error 读取错误
+func (r *tiffBitOrderInput) ReadAt(data []byte, offset int64) (int, error) {
+	n, err := r.source.ReadAt(data, offset)
+	index := sort.Search(len(r.ranges), func(i int) bool { return r.ranges[i][1] > offset })
+	for ; index < len(r.ranges) && r.ranges[index][0] < offset+int64(n); index++ {
+		span := r.ranges[index]
+		for i := max(int64(0), span[0]-offset); i < min(int64(n), span[1]-offset); i++ {
+			data[i] = bits.Reverse8(data[i])
+		}
+	}
+	return n, err
+}
diff --git a/ofdgo_image_tiff_fax.go b/ofdgo_image_tiff_fax.go
new file mode 100644
index 0000000..2b76754
--- /dev/null
+++ b/ofdgo_image_tiff_fax.go
@@ -0,0 +1,185 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"context"
+	"encoding/binary"
+	"fmt"
+	"image"
+	"image/color"
+	"math/bits"
+
+	"github.com/xiaoqidun/pdfgo"
+)
+
+// tiffFaxImage 组合独立解码的传真条带或瓦片，保留紧凑的一位像素存储
+type tiffFaxImage struct {
+	blocks []image.Image
+	bounds image.Rectangle
+	width  int
+	height int
+	across int
+}
+
+// decodeTIFFFax 按TIFF传真选项解码各图像块，复用PDFGo的T.4与T.6解码器
+// 入参: ctx 取消上下文，可为nil, data TIFF编码, layout 已校验的主图布局
+// 返回: image.Image 图像, error 字段、解码或取消错误
+func decodeTIFFFax(ctx context.Context, data []byte, layout tiffImageLayout) (image.Image, error) {
+	if ctx == nil {
+		ctx = context.Background()
+	}
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	if layout.samples != 1 || layout.compression == 3 && layout.t4 & ^uint32(7) != 0 || layout.compression == 4 && layout.t6 & ^uint32(2) != 0 {
+		return nil, fmt.Errorf("invalid TIFF fax options")
+	}
+	var order binary.ByteOrder = binary.LittleEndian
+	if data[0] == 'M' {
+		order = binary.BigEndian
+	}
+	offset := int(layout.offset)
+	count := int(order.Uint16(data[offset:]))
+	entries := make(map[uint16][]byte, count)
+	for i := range count {
+		entry := data[offset+2+i*12 : offset+14+i*12]
+		entries[order.Uint16(entry)] = entry
+	}
+	value := func(tag uint16, fallback uint32) (uint32, error) {
+		entry := entries[tag]
+		if entry == nil {
+			return fallback, nil
+		}
+		if order.Uint32(entry[4:]) != 1 {
+			return 0, fmt.Errorf("invalid TIFF scalar field %d", tag)
+		}
+		return tiffPlaneValue(data, order, entry, 0)
+	}
+	for _, tag := range []uint16{258, 317, 339} {
+		v, err := value(tag, 1)
+		if err != nil || v != 1 {
+			return nil, fmt.Errorf("invalid TIFF fax sample field %d", tag)
+		}
+	}
+	photo, err := value(262, 2)
+	if err != nil || photo > 1 || entries[338] != nil {
+		return nil, fmt.Errorf("invalid TIFF fax color model")
+	}
+	width := uint32(layout.width)
+	height, err := value(278, ^uint32(0))
+	if err != nil || height == 0 {
+		return nil, fmt.Errorf("invalid TIFF rows per strip")
+	}
+	height = min(height, uint32(layout.height))
+	locations, lengths := entries[273], entries[279]
+	tiled := entries[322] != nil
+	if tiled {
+		width, err = value(322, 0)
+		if err != nil || width == 0 || width%16 != 0 {
+			return nil, fmt.Errorf("invalid TIFF tile width")
+		}
+		height, err = value(323, 0)
+		if err != nil || height == 0 || height%16 != 0 {
+			return nil, fmt.Errorf("invalid TIFF tile height")
+		}
+		locations, lengths = entries[324], entries[325]
+		if locations != nil && order.Uint16(locations[2:]) != 4 {
+			return nil, fmt.Errorf("invalid TIFF tile offsets")
+		}
+	}
+	if uint64(width) > uint64(int(^uint(0)>>1)) || uint64(height) > uint64(int(^uint(0)>>1)) || locations == nil || lengths == nil {
+		return nil, fmt.Errorf("invalid TIFF fax blocks")
+	}
+	across := (uint64(layout.width)-1)/uint64(width) + 1
+	down := (uint64(layout.height)-1)/uint64(height) + 1
+	blocks := across * down
+	if blocks != uint64(order.Uint32(locations[4:])) || blocks != uint64(order.Uint32(lengths[4:])) {
+		return nil, fmt.Errorf("invalid TIFF fax block count")
+	}
+	result := &tiffFaxImage{bounds: image.Rect(0, 0, layout.width, layout.height), width: int(width), height: int(height), across: int(across)}
+	for i := uint64(0); i < blocks; i++ {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
+		at, err := tiffPlaneValue(data, order, locations, uint32(i))
+		if err != nil {
+			return nil, err
+		}
+		size, err := tiffPlaneValue(data, order, lengths, uint32(i))
+		if err != nil || size == 0 || uint64(at)+uint64(size) > uint64(len(data)) {
+			return nil, fmt.Errorf("truncated TIFF fax block")
+		}
+		encoded := data[uint64(at) : uint64(at)+uint64(size)]
+		if layout.fillOrder == 2 {
+			reversed := make([]byte, len(encoded))
+			for start := 0; start < len(encoded); {
+				if err := ctx.Err(); err != nil {
+					return nil, err
+				}
+				end := start + min(32768, len(encoded)-start)
+				for j := start; j < end; j++ {
+					reversed[j] = bits.Reverse8(encoded[j])
+				}
+				start = end
+			}
+			encoded = reversed
+		}
+		rows := height
+		if !tiled {
+			rows = min(height, uint32(uint64(layout.height)-i*uint64(height)))
+		}
+		k := pdfgo.Integer(0)
+		if layout.compression == 4 {
+			k = -1
+		} else if layout.compression == 3 && layout.t4&1 != 0 {
+			k = 1
+		}
+		params := pdfgo.Dictionary{
+			"K": k, "Columns": pdfgo.Integer(width), "Rows": pdfgo.Integer(rows),
+			"EndOfBlock": pdfgo.Boolean(false), "EndOfLine": pdfgo.Boolean(layout.compression == 3),
+			"EncodedByteAlign": pdfgo.Boolean(layout.compression == 2), "BlackIs1": pdfgo.Boolean(photo == 1),
+		}
+		object := &pdfgo.Image{Width: int(width), Height: int(rows), BitsPerComponent: 1, ColorSpace: pdfgo.Name("DeviceGray"), Stream: &pdfgo.Stream{Dictionary: pdfgo.Dictionary{"Filter": pdfgo.Name("CCITTFaxDecode"), "DecodeParms": params}, Data: encoded}}
+		pixels, err := object.DecodeSamplesContext(ctx)
+		if err != nil {
+			return nil, fmt.Errorf("TIFF fax block %d: %w", i, err)
+		}
+		result.blocks = append(result.blocks, pixels)
+	}
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	return result, nil
+}
+
+// Bounds 返回主图显示区域，不包含边缘瓦片的填充像素
+// 返回: image.Rectangle 图像区域
+func (s *tiffFaxImage) Bounds() image.Rectangle { return s.bounds }
+
+// ColorModel 返回二值图像使用的灰度颜色模型
+// 返回: color.Model 灰度模型
+func (s *tiffFaxImage) ColorModel() color.Model { return color.GrayModel }
+
+// At 读取所属条带或瓦片中的像素，区域外返回黑色
+// 入参: x 横向像素坐标, y 纵向像素坐标
+// 返回: color.Color 像素颜色
+func (s *tiffFaxImage) At(x, y int) color.Color {
+	if !image.Pt(x, y).In(s.bounds) {
+		return color.Gray{}
+	}
+	block := s.blocks[(y/s.height)*s.across+x/s.width]
+	return block.At(x%s.width, y%s.height)
+}
diff --git a/ofdgo_image_tiff_planar.go b/ofdgo_image_tiff_planar.go
new file mode 100644
index 0000000..9b83063
--- /dev/null
+++ b/ofdgo_image_tiff_planar.go
@@ -0,0 +1,527 @@
+// Copyright 2025-2026 肖其顿 (XIAO QI DUN)
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ofdgo
+
+import (
+	"bytes"
+	"context"
+	"encoding/binary"
+	"fmt"
+	"image"
+	"image/color"
+	"io"
+	"math"
+)
+
+// tiffPlaneInput 将单分量目录附加到只读视图，像素仍从原始编码读取
+type tiffPlaneInput struct {
+	source, directory []byte
+	position          int64
+	order             binary.ByteOrder
+	ctx               context.Context
+}
+
+// planarTIFFImage 组合独立分量或映射调色板，不额外分配交错像素缓冲
+type planarTIFFImage struct {
+	planes  []image.Image
+	bounds  image.Rectangle
+	depth   uint32
+	packed  [4]uint32
+	alpha   uint32
+	gray    bool
+	palette []color.NRGBA64
+}
+
+// decodeTIFFImage 按存储方式解码主图并统一显示方向
+// 入参: ctx 取消上下文，可为nil, data TIFF编码, layout 主图布局
+// 返回: image.Image 图像, error 解码或取消错误
+func decodeTIFFImage(ctx context.Context, data []byte, layout tiffImageLayout) (image.Image, error) {
+	if layout.compression >= 2 && layout.compression <= 4 {
+		pixels, err := decodeTIFFFax(ctx, data, layout)
+		if err == nil && layout.orientation != 1 {
+			pixels = &orientedTIFFImage{source: pixels, tiffImageLayout: layout}
+		}
+		return pixels, err
+	}
+	ranges, err := tiffBitOrderRanges(ctx, data, layout)
+	if err != nil {
+		return nil, err
+	}
+	var pixels image.Image
+	if layout.planar && layout.samples > 1 || layout.samples == 1 && (layout.depth == 2 || layout.depth == 4) {
+		result, directories, err := tiffPlaneDirectories(ctx, data, layout)
+		if err != nil {
+			return nil, err
+		}
+		for _, directory := range directories {
+			input := &tiffPlaneInput{source: data, directory: directory, order: binary.LittleEndian, ctx: ctx}
+			if data[0] == 'M' {
+				input.order = binary.BigEndian
+			}
+			plane, err := decodeTIFFBitOrder(input, ranges)
+			if err != nil {
+				return nil, err
+			}
+			result.planes = append(result.planes, plane)
+		}
+		pixels = result
+	} else {
+		input := newTIFFDirectoryInput(data, layout.offset)
+		input.ctx = ctx
+		var err error
+		pixels, err = decodeTIFFBitOrder(input, ranges)
+		if err != nil {
+			return nil, err
+		}
+	}
+	if ctx != nil {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
+	}
+	if layout.orientation != 1 {
+		pixels = &orientedTIFFImage{source: pixels, tiffImageLayout: layout}
+	}
+	return pixels, nil
+}
+
+// tiffPlaneDirectories 为颜色、透明度或调色板索引构建单通道TIFF目录
+// 入参: ctx 取消上下文，可为nil, data TIFF编码, layout 已校验的主图布局
+// 返回: *planarTIFFImage 像素布局, [][]byte 各分量目录, error 字段或布局错误
+func tiffPlaneDirectories(ctx context.Context, data []byte, layout tiffImageLayout) (*planarTIFFImage, [][]byte, error) {
+	if ctx != nil {
+		if err := ctx.Err(); err != nil {
+			return nil, nil, err
+		}
+	}
+	if uint64(len(data))+65536*12+8 > math.MaxUint32 {
+		return nil, nil, fmt.Errorf("TIFF plane directory offset exceeds format limit")
+	}
+	var order binary.ByteOrder = binary.LittleEndian
+	if data[0] == 'M' {
+		order = binary.BigEndian
+	}
+	offset := int(layout.offset)
+	count := int(order.Uint16(data[offset:]))
+	entries := make(map[uint16][]byte, count)
+	for i := range count {
+		entry := data[offset+2+i*12 : offset+14+i*12]
+		entries[order.Uint16(entry)] = entry
+	}
+	value := func(tag uint16, fallback uint32) (uint32, error) {
+		entry := entries[tag]
+		if entry == nil {
+			return fallback, nil
+		}
+		if order.Uint32(entry[4:]) != 1 {
+			return 0, fmt.Errorf("invalid TIFF scalar field %d", tag)
+		}
+		return tiffPlaneValue(data, order, entry, 0)
+	}
+	samples, err := value(277, 1)
+	if err != nil || samples < 1 || samples > 4 {
+		return nil, nil, fmt.Errorf("invalid TIFF plane count")
+	}
+	photo, err := value(262, math.MaxUint32)
+	if err != nil || photo > 3 || photo == 3 && samples != 1 {
+		return nil, nil, fmt.Errorf("unsupported TIFF planar color model")
+	}
+	compression, err := value(259, 1)
+	if err != nil {
+		return nil, nil, err
+	}
+	fill, err := value(266, 1)
+	if err != nil || fill != 1 && (fill != 2 || compression != 1 && compression != 3 && compression != 4) {
+		return nil, nil, fmt.Errorf("unsupported TIFF plane fill order")
+	}
+	predictor, err := value(317, 1)
+	if err != nil || predictor != 1 && predictor != 2 {
+		return nil, nil, fmt.Errorf("unsupported TIFF plane predictor")
+	}
+	result := &planarTIFFImage{bounds: image.Rect(0, 0, layout.width, layout.height), gray: photo < 2}
+	colors := uint32(3)
+	if photo != 2 {
+		colors = 1
+	}
+	if samples != colors {
+		result.alpha, err = value(338, 0)
+		if err != nil || samples != colors+1 || result.alpha != 1 && result.alpha != 2 {
+			return nil, nil, fmt.Errorf("unsupported TIFF extra planes")
+		}
+	} else if entries[338] != nil {
+		return nil, nil, fmt.Errorf("unexpected TIFF extra samples")
+	}
+	width, err := value(322, 0)
+	if err != nil {
+		return nil, nil, err
+	}
+	blocks := uint64(0)
+	locations, lengths := uint16(273), uint16(279)
+	if width != 0 {
+		height, err := value(323, 0)
+		if err != nil || height == 0 || width%16 != 0 || height%16 != 0 {
+			return nil, nil, fmt.Errorf("invalid TIFF tile dimensions")
+		}
+		blocks = ((uint64(layout.width)-1)/uint64(width) + 1) * ((uint64(layout.height)-1)/uint64(height) + 1)
+		locations, lengths = 324, 325
+	} else {
+		rows, err := value(278, math.MaxUint32)
+		if err != nil || rows == 0 {
+			return nil, nil, fmt.Errorf("invalid TIFF rows per strip")
+		}
+		blocks = (uint64(layout.height)-1)/uint64(rows) + 1
+	}
+	if blocks > math.MaxUint32/uint64(samples) {
+		return nil, nil, fmt.Errorf("too many TIFF plane blocks")
+	}
+	for _, tag := range []uint16{locations, lengths} {
+		entry := entries[tag]
+		if entry == nil || uint64(order.Uint32(entry[4:])) != blocks*uint64(samples) || tag == 324 && order.Uint16(entry[2:]) != 4 {
+			return nil, nil, fmt.Errorf("invalid TIFF plane block count")
+		}
+	}
+	for i := uint32(0); i < uint32(blocks)*samples; i++ {
+		if ctx != nil {
+			if err := ctx.Err(); err != nil {
+				return nil, nil, err
+			}
+		}
+		at, err := tiffPlaneValue(data, order, entries[locations], i)
+		if err != nil {
+			return nil, nil, err
+		}
+		size, err := tiffPlaneValue(data, order, entries[lengths], i)
+		if err != nil || size == 0 || uint64(at)+uint64(size) > uint64(len(data)) {
+			return nil, nil, fmt.Errorf("truncated TIFF plane block")
+		}
+	}
+	for _, tag := range []uint16{258, 339} {
+		if entry := entries[tag]; entry != nil && (order.Uint16(entry[2:]) != 3 || order.Uint32(entry[4:]) != samples) {
+			return nil, nil, fmt.Errorf("invalid TIFF sample array")
+		}
+	}
+	var directories [][]byte
+	for plane := uint32(0); plane < samples; plane++ {
+		depth := uint32(1)
+		if entry := entries[258]; entry != nil {
+			depth, err = tiffPlaneValue(data, order, entry, plane)
+		}
+		if err != nil || depth != 1 && depth != 2 && depth != 4 && depth != 8 && depth != 16 {
+			return nil, nil, fmt.Errorf("unsupported TIFF plane depth")
+		}
+		result.depth = max(result.depth, depth)
+		if depth < 8 && predictor != 1 {
+			return nil, nil, fmt.Errorf("unsupported TIFF packed plane predictor")
+		}
+		if depth == 2 || depth == 4 {
+			result.packed[plane] = depth
+			if uint64(layout.width)*uint64(depth) > uint64(math.MaxInt)/uint64(layout.height) || uint64(layout.width)*uint64(depth) > math.MaxUint32 || uint64(width)*uint64(depth) > math.MaxUint32 {
+				return nil, nil, fmt.Errorf("TIFF plane dimensions exceed format limit")
+			}
+			if compression == 2 || compression == 3 || compression == 4 {
+				return nil, nil, fmt.Errorf("unsupported TIFF packed plane compression")
+			}
+		}
+		if entry := entries[339]; entry != nil {
+			format, err := tiffPlaneValue(data, order, entry, plane)
+			if err != nil || format != 1 {
+				return nil, nil, fmt.Errorf("unsupported TIFF sample format")
+			}
+		}
+		if photo == 3 {
+			entry := entries[320]
+			if depth > 8 || entry == nil || order.Uint16(entry[2:]) != 3 || order.Uint32(entry[4:]) != 3<<depth {
+				return nil, nil, fmt.Errorf("invalid TIFF color map")
+			}
+			n := uint32(1) << depth
+			result.palette = make([]color.NRGBA64, n)
+			for i := range n {
+				var components [3]uint16
+				for c := range components {
+					v, err := tiffPlaneValue(data, order, entry, uint32(c)*n+i)
+					if err != nil {
+						return nil, nil, err
+					}
+					components[c] = uint16(v)
+					if v%257 != 0 {
+						result.depth = 16
+					}
+				}
+				result.palette[i] = color.NRGBA64{R: components[0], G: components[1], B: components[2], A: 65535}
+			}
+		}
+		directory := make([]byte, 2, 6+count*12)
+		for i := range count {
+			if ctx != nil {
+				if err := ctx.Err(); err != nil {
+					return nil, nil, err
+				}
+			}
+			entry := bytes.Clone(data[offset+2+i*12 : offset+14+i*12])
+			tag := order.Uint16(entry)
+			if tag == 338 || tag == 320 {
+				continue
+			}
+			scalar := uint32(0)
+			switch tag {
+			case 258:
+				scalar = depth
+				if result.packed[plane] != 0 {
+					scalar = 1
+				}
+			case 256, 322:
+				if result.packed[plane] != 0 {
+					original, err := tiffPlaneValue(data, order, entry, 0)
+					if err != nil {
+						return nil, nil, err
+					}
+					order.PutUint16(entry[2:], 4)
+					order.PutUint32(entry[4:], 1)
+					order.PutUint32(entry[8:], original*depth)
+				}
+			case 262:
+				scalar = 1
+				if result.gray && plane == 0 && photo == 0 {
+					scalar = 0
+				}
+			case 277, 284, 339:
+				scalar = 1
+			case locations, lengths:
+				entry, err = tiffPlaneArray(data, order, entry, uint32(blocks)*plane, uint32(blocks))
+				if err != nil {
+					return nil, nil, err
+				}
+			}
+			if tag == 258 || tag == 262 || tag == 277 || tag == 284 || tag == 339 {
+				order.PutUint16(entry[2:], 3)
+				order.PutUint32(entry[4:], 1)
+				clear(entry[8:])
+				order.PutUint16(entry[8:], uint16(scalar))
+			}
+			directory = append(directory, entry...)
+		}
+		order.PutUint16(directory, uint16((len(directory)-2)/12))
+		directories = append(directories, append(directory, 0, 0, 0, 0))
+	}
+	return result, directories, nil
+}
+
+// tiffPlaneValue 读取SHORT或LONG数组中的一项，校验完整数组范围
+// 入参: data 原编码, order 字节序, entry 目录项, index 元素序号
+// 返回: uint32 元素值, error 类型或越界错误
+func tiffPlaneValue(data []byte, order binary.ByteOrder, entry []byte, index uint32) (uint32, error) {
+	kind, count := order.Uint16(entry[2:]), order.Uint32(entry[4:])
+	if kind != 3 && kind != 4 || index >= count {
+		return 0, fmt.Errorf("invalid TIFF plane array")
+	}
+	size := uint64(2)
+	if kind == 4 {
+		size = 4
+	}
+	values := entry[8:]
+	if size*uint64(count) > 4 {
+		at := uint64(order.Uint32(values))
+		if at+size*uint64(count) > uint64(len(data)) {
+			return 0, fmt.Errorf("truncated TIFF plane array")
+		}
+		values = data[at:]
+	}
+	if size == 2 {
+		return uint32(order.Uint16(values[uint64(index)*size:])), nil
+	}
+	return order.Uint32(values[uint64(index)*size:]), nil
+}
+
+// tiffPlaneArray 将块索引缩减为指定分量的范围，较长数组继续借用原编码
+// 入参: data 原编码, order 字节序, entry 目录项副本, first 起始序号, count 元素数
+// 返回: []byte 修改后的目录项, error 类型或越界错误
+func tiffPlaneArray(data []byte, order binary.ByteOrder, entry []byte, first, count uint32) ([]byte, error) {
+	if _, err := tiffPlaneValue(data, order, entry, first+count-1); err != nil {
+		return nil, err
+	}
+	size := uint32(2)
+	if order.Uint16(entry[2:]) == 4 {
+		size = 4
+	}
+	if uint64(count)*uint64(size) <= 4 {
+		var values [4]byte
+		for i := uint32(0); i < count; i++ {
+			value, _ := tiffPlaneValue(data, order, entry, first+i)
+			if size == 2 {
+				order.PutUint16(values[i*size:], uint16(value))
+			} else {
+				order.PutUint32(values[i*size:], value)
+			}
+		}
+		copy(entry[8:], values[:])
+	} else {
+		order.PutUint32(entry[8:], order.Uint32(entry[8:])+first*size)
+	}
+	order.PutUint32(entry[4:], count)
+	return entry, nil
+}
+
+// Read 从组合视图顺序读取
+// 入参: data 接收缓冲
+// 返回: int 字节数, error 读取或取消错误
+func (r *tiffPlaneInput) Read(data []byte) (int, error) {
+	n, err := r.ReadAt(data, r.position)
+	r.position += int64(n)
+	return n, err
+}
+
+// ReadAt 从原始编码或附加目录读取并替换首目录指针
+// 入参: data 接收缓冲, offset 文件偏移
+// 返回: int 字节数, error 读取或取消错误
+func (r *tiffPlaneInput) ReadAt(data []byte, offset int64) (int, error) {
+	if r.ctx != nil {
+		if err := r.ctx.Err(); err != nil {
+			return 0, err
+		}
+	}
+	if offset < 0 {
+		return 0, fmt.Errorf("negative TIFF read offset")
+	}
+	start := int64(len(r.source))
+	directory := start + start%2
+	written := 0
+	if offset < start {
+		written = copy(data, r.source[offset:])
+	}
+	if offset+int64(written) == start && start != directory && written < len(data) {
+		data[written] = 0
+		written++
+	}
+	if at := offset + int64(written) - directory; at >= 0 && at < int64(len(r.directory)) {
+		written += copy(data[written:], r.directory[at:])
+	}
+	var pointer [4]byte
+	r.order.PutUint32(pointer[:], uint32(directory))
+	for i, value := range pointer {
+		if at := int64(4+i) - offset; at >= 0 && at < int64(written) {
+			data[at] = value
+		}
+	}
+	if written != len(data) {
+		return written, io.EOF
+	}
+	return written, nil
+}
+
+// Bounds 返回存储坐标范围
+// 返回: image.Rectangle 图像范围
+func (p *planarTIFFImage) Bounds() image.Rectangle { return p.bounds }
+
+// ColorModel 按透明度类型和最高分量精度返回颜色模型
+// 返回: color.Model 颜色模型
+func (p *planarTIFFImage) ColorModel() color.Model {
+	if p.gray && p.alpha == 0 {
+		if p.depth > 8 {
+			return color.Gray16Model
+		}
+		return color.GrayModel
+	}
+	if p.alpha == 1 {
+		if p.depth > 8 {
+			return color.RGBA64Model
+		}
+		return color.RGBAModel
+	}
+	if p.depth > 8 {
+		return color.NRGBA64Model
+	}
+	return color.NRGBAModel
+}
+
+// components 直接读取灰度分量，保留原始透明度语义
+// 入参: x 横坐标, y 纵坐标
+// 返回: [4]uint16 红、绿、蓝和透明度分量
+func (p *planarTIFFImage) components(x, y int) [4]uint16 {
+	if !image.Pt(x, y).In(p.bounds) {
+		return [4]uint16{}
+	}
+	values := [4]uint16{0, 0, 0, 65535}
+	for i, plane := range p.planes {
+		var value uint16
+		switch plane := plane.(type) {
+		case *image.Gray:
+			if bits := p.packed[i]; bits != 0 {
+				for bit := range int(bits) {
+					value = value<<1 | uint16(plane.GrayAt(x*int(bits)+bit, y).Y>>7)
+				}
+				value *= 65535 / uint16(1<<bits-1)
+			} else {
+				value = uint16(plane.GrayAt(x, y).Y) * 257
+			}
+		case *image.Gray16:
+			value = plane.Gray16At(x, y).Y
+		default:
+			v, _, _, _ := plane.At(x, y).RGBA()
+			value = uint16(v)
+		}
+		if len(p.palette) != 0 {
+			c := p.palette[uint32(value)*uint32(len(p.palette)-1)/65535]
+			return [4]uint16{c.R, c.G, c.B, c.A}
+		}
+		if p.gray && i == 0 {
+			values[0], values[1], values[2] = value, value, value
+		} else if p.gray {
+			values[3] = value
+		} else {
+			values[i] = value
+		}
+	}
+	return values
+}
+
+// At 返回组合后的颜色，关联透明度保留预乘语义
+// 入参: x 横坐标, y 纵坐标
+// 返回: color.Color 像素颜色
+func (p *planarTIFFImage) At(x, y int) color.Color {
+	values := p.components(x, y)
+	if p.alpha == 1 {
+		return color.RGBA64{R: values[0], G: values[1], B: values[2], A: values[3]}
+	}
+	return color.NRGBA64{R: values[0], G: values[1], B: values[2], A: values[3]}
+}
+
+// NRGBA64At 读取独立透明度颜色，保留非预乘分量中的隐藏颜色
+// 入参: x 横坐标, y 纵坐标
+// 返回: color.NRGBA64 像素颜色
+func (p *planarTIFFImage) NRGBA64At(x, y int) color.NRGBA64 {
+	values := p.components(x, y)
+	if p.alpha == 1 && values[3] != 65535 {
+		if values[3] == 0 {
+			return color.NRGBA64{}
+		}
+		for i := range 3 {
+			values[i] = uint16(uint32(values[i]) * 65535 / uint32(values[3]))
+		}
+	}
+	return color.NRGBA64{R: values[0], G: values[1], B: values[2], A: values[3]}
+}
+
+// RGBA64At 返回预乘十六位颜色，不分配临时颜色对象
+// 入参: x 横坐标, y 纵坐标
+// 返回: color.RGBA64 像素颜色
+func (p *planarTIFFImage) RGBA64At(x, y int) color.RGBA64 {
+	values := p.components(x, y)
+	if p.alpha != 1 && values[3] != 65535 {
+		for i := range 3 {
+			values[i] = uint16(uint32(values[i]) * uint32(values[3]) / 65535)
+		}
+	}
+	return color.RGBA64{R: values[0], G: values[1], B: values[2], A: values[3]}
+}
diff --git a/ofdgo_reader.go b/ofdgo_reader.go
index af618ed..103c3a3 100644
--- a/ofdgo_reader.go
+++ b/ofdgo_reader.go
@@ -190,7 +190,7 @@ func (r *Reader) docStructure() (*Document, error) {
 // 入参: page 页面对象
 // 返回: *PageContent 页面内容, error 错误信息
 func (r *Reader) PageContent(page Page) (*PageContent, error) {
-	return r.readPageContent(page, false)
+	return r.readPageContent(page, false, false)
 }
 
 // PageCount 获取当前文档总页数
@@ -729,9 +729,9 @@ func resolveResourcePath(resPath, baseLoc, filePath string) string {
 }
 
 // readPageContent 读取页面，可跳过文字提取不使用的独立路径和图片
-// 入参: page 页面对象, textOnly 是否仅用于文字提取
+// 入参: page 页面对象, textOnly 是否仅用于文字提取, template 是否按模板页读取
 // 返回: *PageContent 页面内容, error 错误信息
-func (r *Reader) readPageContent(page Page, textOnly bool) (*PageContent, error) {
+func (r *Reader) readPageContent(page Page, textOnly, template bool) (*PageContent, error) {
 	fullPath := r.ResPath(page.BaseLoc)
 	data, err := r.readFile(fullPath)
 	if err != nil {
@@ -748,6 +748,9 @@ func (r *Reader) readPageContent(page Page, textOnly bool) (*PageContent, error)
 		return nil, fmt.Errorf("failed to unmarshal page content: %w", err)
 	}
 	content.ID = page.ID
+	if template {
+		content.Template = nil
+	}
 	for _, res := range content.PageRes {
 		r.loadRes(resolveResourcePath(page.BaseLoc, "", res))
 	}
@@ -836,7 +839,7 @@ func (r *Reader) resolvePageArea(area PageArea, templates []Template) (PageArea,
 	}
 	for _, ref := range templates {
 		for _, template := range r.doc.CommonData.TemplatePage {
-			if template.ID != ref.TemplateID {
+			if !sameResourceID(template.ID, ref.TemplateID) {
 				continue
 			}
 			content, err := r.readPageHeader(Page{BaseLoc: template.BaseLoc})
diff --git a/ofdgo_reader_dependencies.go b/ofdgo_reader_dependencies.go
index d5d58a1..537777c 100644
--- a/ofdgo_reader_dependencies.go
+++ b/ofdgo_reader_dependencies.go
@@ -127,27 +127,34 @@ func packageReference(root, parent, name, attribute string) int {
 	return packageReferenceNone
 }
 
-// documentFiles 遍历独立文档及版本依赖，只解析结构文件，不读取附件等原始数据
-// 入参: ctx 取消上下文, index 文档索引
+// documentFiles 汇总指定文档及版本依赖，共享结构只解析一次，不读取附件等原始数据
+// 入参: ctx 取消上下文, indexes 文档索引，为空时不扫描任何文档
 // 返回: map[string]bool 包内文件路径, error 读取或取消错误
-func (r *Reader) documentFiles(ctx context.Context, index int) (map[string]bool, error) {
-	files, _, err := r.documentDependencies(ctx, index)
+func (r *Reader) documentFiles(ctx context.Context, indexes ...int) (map[string]bool, error) {
+	files, _, err := r.documentDependencies(ctx, indexes...)
 	return files, err
 }
 
-// documentDependencies 区分文档全部依赖与需要解析的结构文件，避免重复读取原始数据
-// 入参: ctx 取消上下文, index 文档索引
+// documentDependencies 合并指定入口的依赖图，区分结构文件与原始数据
+// 入参: ctx 取消上下文, indexes 文档索引，为空时返回空集合
 // 返回: map[string]bool 全部依赖, map[string]bool 结构文件, error 读取或取消错误
-func (r *Reader) documentDependencies(ctx context.Context, index int) (map[string]bool, map[string]bool, error) {
+func (r *Reader) documentDependencies(ctx context.Context, indexes ...int) (map[string]bool, map[string]bool, error) {
 	if err := ctx.Err(); err != nil {
 		return nil, nil, err
 	}
-	if index < 0 || index >= r.DocumentCount() {
-		return nil, nil, fmt.Errorf("document index out of range: %d", index)
+	selected := make(map[int]bool, len(indexes))
+	for _, index := range indexes {
+		if index < 0 || index >= r.DocumentCount() {
+			return nil, nil, fmt.Errorf("document index out of range: %d", index)
+		}
+		selected[index] = true
 	}
 	var queue []string
 	seen := make(map[string]bool)
 	queued := make(map[string]bool)
+	if len(selected) == 0 {
+		return seen, queued, ctx.Err()
+	}
 	add := func(name string, kind int) {
 		seen[name] = true
 		if kind == packageReferenceXML && !queued[name] {
@@ -155,22 +162,21 @@ func (r *Reader) documentDependencies(ctx context.Context, index int) (map[strin
 			queue = append(queue, name)
 		}
 	}
-	data, err := r.readFile("OFD.xml")
+	data, err := r.readFileViewContext(ctx, "OFD.xml")
 	if err != nil {
 		return nil, nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return nil, nil, err
 	}
-	body := root.childAt("DocBody", index)
-	if body == nil {
-		return nil, nil, fmt.Errorf("document body not found: %d", index)
-	}
-	var collect func(*editorXML)
-	collect = func(node *editorXML) {
+	var collect func(*editorXML) error
+	collect = func(node *editorXML) error {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
 		if !node.matchesOFD(node.name.Local) {
-			return
+			return nil
 		}
 		parent := node.parent.name.Local
 		if kind := packageReference("OFD", parent, node.name.Local, ""); kind != packageReferenceNone {
@@ -187,10 +193,28 @@ func (r *Reader) documentDependencies(ctx context.Context, index int) (map[strin
 			}
 		}
 		for _, child := range node.children {
-			collect(child)
+			if err := collect(child); err != nil {
+				return err
+			}
 		}
+		return nil
+	}
+	index := 0
+	for _, body := range root.children {
+		if !body.matchesOFD("DocBody") {
+			continue
+		}
+		if selected[index] {
+			if err := collect(body); err != nil {
+				return nil, nil, err
+			}
+			delete(selected, index)
+		}
+		index++
+	}
+	if len(selected) != 0 {
+		return nil, nil, fmt.Errorf("document body not found")
 	}
-	collect(body)
 	for len(queue) > 0 {
 		if err := ctx.Err(); err != nil {
 			return nil, nil, err
diff --git a/ofdgo_render_bounds.go b/ofdgo_render_bounds.go
index 4dc3559..83cddf0 100644
--- a/ofdgo_render_bounds.go
+++ b/ofdgo_render_bounds.go
@@ -75,6 +75,9 @@ func (r *Renderer) AnnotationGeometry(annotation Annotation) (Box, []ObjectConto
 // 入参: object 图形对象, state 绘制状态
 // 返回: error 度量错误
 func (m *annotationMeasurement) DrawObject(object *GraphicObject, state RenderState) error {
+	if state.Hidden {
+		return nil
+	}
 	result, err := m.renderer.MeasureObject(*object, MeasureOptions{Defaults: state.Defaults, Parent: state.Parent, BoundaryInCTM: state.BoundaryInCTM, Clip: state.Clip, Contours: true})
 	if err != nil {
 		return err
diff --git a/ofdgo_render_compile.go b/ofdgo_render_compile.go
index 85ba9bc..e48e76f 100644
--- a/ofdgo_render_compile.go
+++ b/ofdgo_render_compile.go
@@ -44,6 +44,9 @@ type semanticStyle struct {
 // 入参: object 源对象, state 继承状态
 // 返回: error 编译错误
 func (c *semanticCompiler) DrawObject(object *GraphicObject, state RenderState) error {
+	if state.Hidden {
+		return nil
+	}
 	switch object.Type {
 	case "TextObject":
 		return c.textObject(object.TextObject, state)
diff --git a/ofdgo_render_text.go b/ofdgo_render_text.go
index 11276b3..03d1dd0 100644
--- a/ofdgo_render_text.go
+++ b/ofdgo_render_text.go
@@ -49,7 +49,7 @@ type glyphOutlineKey struct {
 func (r *Renderer) PositionText(object TextObject, state RenderState) (*PositionedText, error) {
 	result := &PositionedText{Run: TextRun{ID: object.ID, Text: object.Text()}}
 	result.Run.Boxes = make([]Box, len([]rune(result.Run.Text)))
-	if object.Visible != nil && !*object.Visible {
+	if state.Hidden || object.Visible != nil && !*object.Visible {
 		return result, nil
 	}
 	geometry, err := r.Geometry()
diff --git a/ofdgo_render_walk.go b/ofdgo_render_walk.go
index 308d575..a75593c 100644
--- a/ofdgo_render_walk.go
+++ b/ofdgo_render_walk.go
@@ -17,11 +17,13 @@ package ofdgo
 import "fmt"
 
 // RenderState 保存对象继承的样式、父变换和页面坐标裁剪，所有引用只读
-// Clip为nil表示未裁剪，空路径表示完全裁去，单位为毫米
+// Clip为nil表示未提供裁剪，空路径表示完全裁去，单位为毫米
+// Hidden表示当前对象或其上层容器不可见，不改变源对象的Visible属性
 type RenderState struct {
 	Defaults      *DrawParam
 	Parent        *Matrix
 	BoundaryInCTM bool
+	Hidden        bool
 	Clip          *GeometryPath
 }
 
@@ -42,6 +44,19 @@ type PageGroups interface {
 	SkipObjects(count int)
 }
 
+// PageComposites 可选地接收展开前的复合图元及矢量资源，不替代成员遍历
+// 状态包含当前图元到页面的完整变换，源对象及状态只读
+type PageComposites interface {
+	VisitComposite(object *CompositeGraphicUnit, state RenderState) error
+}
+
+// PageAnalysis 控制WalkPage和WalkObject对隐藏内容及裁剪的处理，不影响已构建场景的重放
+// 跳过裁剪时RenderState.Clip为空，源对象的裁剪定义不变，叶对象仍由访问器解释
+type PageAnalysis interface {
+	IncludeHidden() bool
+	SkipClips() bool
+}
+
 // WalkPage 按模板、图层、注解和印章顺序访问页面，不依赖任何绘图库
 // 页面背景由调用方处理，文字裁剪可通过几何后端加载字体，图片由访问器按需加载
 // 入参: page 页面内容, visitor 页面访问器
@@ -79,7 +94,7 @@ func (r *Renderer) WalkPage(page *PageContent, visitor PageVisitor) error {
 	}
 	if r.RenderAnnotations {
 		for _, annotation := range r.Reader.Annots[page.ID] {
-			if annotation.Visible != nil && !*annotation.Visible {
+			if annotation.Visible != nil && !*annotation.Visible && !pageIncludesHidden(visitor) {
 				continue
 			}
 			if err := r.walkAnnotation(annotation, visitor); err != nil {
@@ -116,7 +131,7 @@ func (r *Renderer) walkAnnotation(annotation Annotation, visitor PageVisitor) er
 	}
 	box, _ := ParseBox(annotation.Appearance.Boundary)
 	matrix := TranslationMatrix(box.X, box.Y)
-	state := RenderState{Parent: &matrix, BoundaryInCTM: true}
+	state := RenderState{Parent: &matrix, BoundaryInCTM: true, Hidden: annotation.Visible != nil && !*annotation.Visible}
 	for i := range annotation.Appearance.Objects {
 		if err := r.WalkObject(&annotation.Appearance.Objects[i], state, visitor); err != nil {
 			return err
@@ -175,11 +190,21 @@ func (r *Renderer) walkObject(object *GraphicObject, state RenderState, visitor
 			return err
 		}
 	}
+	if pageSkipsClips(visitor) {
+		state.Clip = nil
+	}
 	if groups, ok := visitor.(PageGroups); ok && groups.BeginObject(object) {
 		defer groups.EndObject()
 	}
 	switch object.Type {
-	case "TextObject", "PathObject", "ImageObject":
+	case "TextObject":
+		state.Hidden = state.Hidden || object.TextObject.Visible != nil && !*object.TextObject.Visible
+		return visitor.DrawObject(object, state)
+	case "PathObject":
+		state.Hidden = state.Hidden || object.PathObject.Visible != nil && !*object.PathObject.Visible
+		return visitor.DrawObject(object, state)
+	case "ImageObject":
+		state.Hidden = state.Hidden || object.ImageObject.Visible != nil && !*object.ImageObject.Visible
 		return visitor.DrawObject(object, state)
 	case "CompositeGraphicUnit", "CompositeObject":
 		return r.walkComposite(object.CompositeGraphicUnit, state, visitor, references)
@@ -192,7 +217,8 @@ func (r *Renderer) walkObject(object *GraphicObject, state RenderState, visitor
 // 入参: object 复合图元, state 继承状态, visitor 页面访问器, references 当前资源引用链
 // 返回: error 几何或访问器错误
 func (r *Renderer) walkComposite(object CompositeGraphicUnit, state RenderState, visitor PageVisitor, references map[*CompositeGraphicUnit]bool) error {
-	if object.Visible != nil && !*object.Visible {
+	state.Hidden = state.Hidden || object.Visible != nil && !*object.Visible
+	if state.Hidden && !pageIncludesHidden(visitor) {
 		if groups, ok := visitor.(PageGroups); ok {
 			count, ref, seen := len(object.Objects), object.ResourceID, make(map[*CompositeGraphicUnit]bool)
 			for ref != "" {
@@ -213,33 +239,42 @@ func (r *Renderer) walkComposite(object CompositeGraphicUnit, state RenderState,
 		boundary = state.Parent.Multiply(boundary)
 	}
 	matrix := boundary.Multiply(NewMatrix(object.CTM))
-	var err error
-	if object.Boundary != "" {
-		state.Clip, err = r.geometryExtentClip(box.W, box.H, boundary, state.Clip)
+	if pageSkipsClips(visitor) {
+		state.Clip = nil
+	} else {
+		var err error
+		if object.Boundary != "" {
+			state.Clip, err = r.geometryExtentClip(box.W, box.H, boundary, state.Clip)
+			if err != nil {
+				return err
+			}
+		}
+		state.Clip, err = r.compositeExtentClip(object, matrix, state.Clip)
 		if err != nil {
 			return err
 		}
-	}
-	state.Clip, err = r.compositeExtentClip(object, matrix, state.Clip)
-	if err != nil {
-		return err
-	}
-	clips, clipMatrix := object.Clips, matrix
-	if !clips.usesObjectMatrix() {
-		clipMatrix = boundary
-	}
-	if clips != nil {
-		geometry, err := r.Geometry()
-		if err != nil {
-			return err
+		clips, clipMatrix := object.Clips, matrix
+		if !clips.usesObjectMatrix() {
+			clipMatrix = boundary
 		}
-		state.Clip, err = geometry.Clip(r, clips, clipMatrix, state.Clip)
-		if err != nil {
-			return err
+		if clips != nil {
+			geometry, err := r.Geometry()
+			if err != nil {
+				return err
+			}
+			state.Clip, err = geometry.Clip(r, clips, clipMatrix, state.Clip)
+			if err != nil {
+				return err
+			}
 		}
 	}
 	state.Defaults = r.drawParamDefaults(object.DrawParam, state.Defaults)
 	state.Parent = &matrix
+	if composites, ok := visitor.(PageComposites); ok {
+		if err := composites.VisitComposite(&object, state); err != nil {
+			return err
+		}
+	}
 	if object.ResourceID != "" {
 		if ref, _ := resourceValue(r.CompositeGraphicUnits, object.ResourceID); ref != nil {
 			if references[ref] {
@@ -269,6 +304,22 @@ func (r *Renderer) walkComposite(object CompositeGraphicUnit, state RenderState,
 	return nil
 }
 
+// pageIncludesHidden 查询访问器是否需要分析隐藏内容，不改变渲染器配置
+// 入参: visitor 页面访问器
+// 返回: bool 是否访问隐藏内容
+func pageIncludesHidden(visitor PageVisitor) bool {
+	analysis, ok := visitor.(PageAnalysis)
+	return ok && analysis.IncludeHidden()
+}
+
+// pageSkipsClips 查询访问器是否只需资源和变换，不计算几何裁剪
+// 入参: visitor 页面访问器
+// 返回: bool 是否跳过裁剪
+func pageSkipsClips(visitor PageVisitor) bool {
+	analysis, ok := visitor.(PageAnalysis)
+	return ok && analysis.SkipClips()
+}
+
 // compositeExtentClip 按矢量资源尺寸裁剪内容，资源裁剪随实例变换
 // 入参: object 矢量资源, matrix 资源到页面矩阵, parent 父裁剪
 // 返回: *GeometryPath 页面裁剪, error 尺寸或几何错误
diff --git a/ofdgo_renderer.go b/ofdgo_renderer.go
index 1b0e813..c6f93b2 100644
--- a/ofdgo_renderer.go
+++ b/ofdgo_renderer.go
@@ -113,18 +113,21 @@ func (r *Renderer) childRenderer(reader *Reader) *Renderer {
 	return renderer
 }
 
-// loadTemplate 加载模板页面并复用解析结果
+// loadTemplate 按标识数值加载模板页面，等价引用共用解析结果
 // 入参: templateID 模板ID
 // 返回: *PageContent 模板页面
 func (r *Renderer) loadTemplate(templateID string) *PageContent {
+	if key := editorResourceID(templateID); key != "" {
+		templateID = key
+	}
 	if page := r.templatePageCache[templateID]; page != nil {
 		return page
 	}
 	for _, tpl := range r.Reader.doc.CommonData.TemplatePage {
-		if tpl.ID != templateID {
+		if !sameResourceID(tpl.ID, templateID) {
 			continue
 		}
-		page, err := r.Reader.PageContent(Page{BaseLoc: tpl.BaseLoc})
+		page, err := r.Reader.readPageContent(Page{BaseLoc: tpl.BaseLoc}, false, true)
 		if err != nil {
 			return nil
 		}
diff --git a/ofdgo_renderer_layer_canvas.go b/ofdgo_renderer_layer_canvas.go
index 19f6adb..317b9af 100644
--- a/ofdgo_renderer_layer_canvas.go
+++ b/ofdgo_renderer_layer_canvas.go
@@ -27,6 +27,9 @@ type canvasPageVisitor struct {
 // 入参: object 源对象, state 继承状态
 // 返回: error 绘制错误
 func (v *canvasPageVisitor) DrawObject(object *GraphicObject, state RenderState) error {
+	if state.Hidden {
+		return nil
+	}
 	clip, err := geometryToCanvasPath(state.Clip)
 	if err != nil {
 		return err
diff --git a/ofdgo_search.go b/ofdgo_search.go
index ae28bf8..ce1d282 100644
--- a/ofdgo_search.go
+++ b/ofdgo_search.go
@@ -113,7 +113,7 @@ func (r *Renderer) PageTextByIndex(index int) (*PageText, error) {
 	if index < 0 || index >= len(doc.Pages.Page) {
 		return nil, fmt.Errorf("page index %d out of range", index)
 	}
-	page, err := r.Reader.readPageContent(doc.Pages.Page[index], true)
+	page, err := r.Reader.readPageContent(doc.Pages.Page[index], true, false)
 	if err != nil {
 		return nil, err
 	}
diff --git a/ofdgo_version.go b/ofdgo_version.go
index f174e69..411f6da 100644
--- a/ofdgo_version.go
+++ b/ofdgo_version.go
@@ -15,6 +15,7 @@
 package ofdgo
 
 import (
+	"context"
 	"fmt"
 	"strconv"
 	"strings"
@@ -54,32 +55,53 @@ func (r *Reader) DocumentVersionCount() (int, error) {
 // 返回独立副本；没有版本列表时返回nil，不以版本号推测默认版本
 // 返回: []DocumentVersionInfo 版本信息, error 读取或描述错误
 func (r *Reader) DocumentVersions() ([]DocumentVersionInfo, error) {
-	entries, err := r.documentVersionEntries()
+	return r.DocumentVersionsContext(context.Background())
+}
+
+// DocumentVersionsContext 按入口顺序读取版本信息，支持取消描述文件和文件清单的解析
+// 返回独立副本，不切换当前版本，也不读取文件清单中的内容
+// 入参: ctx 取消上下文
+// 返回: []DocumentVersionInfo 版本信息, error 读取、描述或取消错误
+func (r *Reader) DocumentVersionsContext(ctx context.Context) ([]DocumentVersionInfo, error) {
+	entries, err := r.documentVersionEntriesContext(ctx)
 	if err != nil {
 		return nil, err
 	}
 	for i, entry := range entries {
-		info, err := r.readDocumentVersion(entry.ID, entry.Location)
+		info, err := r.readDocumentVersionContext(ctx, entry.ID, entry.Location)
 		if err != nil {
 			return nil, fmt.Errorf("version %s: %w", entry.ID, err)
 		}
 		info.Index, info.Current = entry.Index, entry.Current
 		entries[i] = info
 	}
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	return entries, nil
 }
 
 // documentVersionEntries 读取当前文档的版本入口，不提前加载各版本描述
 // 返回: []DocumentVersionInfo 入口属性, error 读取或描述错误
 func (r *Reader) documentVersionEntries() ([]DocumentVersionInfo, error) {
+	return r.documentVersionEntriesContext(context.Background())
+}
+
+// documentVersionEntriesContext 读取版本入口，在读取和解析过程中检查取消状态
+// 入参: ctx 取消上下文
+// 返回: []DocumentVersionInfo 入口属性, error 读取、描述或取消错误
+func (r *Reader) documentVersionEntriesContext(ctx context.Context) ([]DocumentVersionInfo, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	if r.documentIndex < 0 || r.documentIndex >= r.DocumentCount() {
 		return nil, fmt.Errorf("no docbody found")
 	}
-	data, err := r.readFile("OFD.xml")
+	data, err := r.readFileViewContext(ctx, "OFD.xml")
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -97,6 +119,9 @@ func (r *Reader) documentVersionEntries() ([]DocumentVersionInfo, error) {
 	var result []DocumentVersionInfo
 	ids := make(map[string]bool)
 	for _, node := range versions.children {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		if !node.matchesOFD("Version") {
 			continue
 		}
@@ -135,12 +160,19 @@ func (r *Reader) documentVersionEntries() ([]DocumentVersionInfo, error) {
 // 入参: id 入口中的版本标识, location 描述文件路径
 // 返回: DocumentVersionInfo 版本信息, error 读取或描述错误
 func (r *Reader) readDocumentVersion(id, location string) (DocumentVersionInfo, error) {
+	return r.readDocumentVersionContext(context.Background(), id, location)
+}
+
+// readDocumentVersionContext 读取版本描述，按描述文件位置解析清单和文档入口
+// 入参: ctx 取消上下文, id 入口中的版本标识, location 描述文件路径
+// 返回: DocumentVersionInfo 版本信息, error 读取、描述或取消错误
+func (r *Reader) readDocumentVersionContext(ctx context.Context, id, location string) (DocumentVersionInfo, error) {
 	var info DocumentVersionInfo
-	data, err := r.readFile(location)
+	data, err := r.readFileViewContext(ctx, location)
 	if err != nil {
 		return info, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return info, err
 	}
@@ -162,6 +194,9 @@ func (r *Reader) readDocumentVersion(id, location string) (DocumentVersionInfo,
 	ids := map[string]bool{id: true}
 	found := false
 	for _, node := range list.children {
+		if err := ctx.Err(); err != nil {
+			return DocumentVersionInfo{}, err
+		}
 		if !node.matchesOFD("File") {
 			continue
 		}
diff --git a/ofdgo_version_edit.go b/ofdgo_version_edit.go
index 0e39d3b..ed6bb39 100644
--- a/ofdgo_version_edit.go
+++ b/ofdgo_version_edit.go
@@ -19,6 +19,7 @@ import (
 	"context"
 	"encoding/xml"
 	"fmt"
+	"io"
 	"maps"
 	"slices"
 	"strconv"
@@ -88,15 +89,13 @@ func (r *Reader) versionFileDependencies(ctx context.Context, entry string, part
 func (r *Reader) versionEditGraph(ctx context.Context, entry string, parts map[string][]byte, generated map[string]editorGeneratedReferences) (*versionEditGraph, error) {
 	graph := &versionEditGraph{files: make(map[string]*versionEditFile), used: make(map[string]bool)}
 	queue := []string{entry}
+	queued := map[string]bool{entry: true}
 	for len(queue) != 0 {
 		if err := ctx.Err(); err != nil {
 			return nil, err
 		}
 		name := queue[0]
 		queue = queue[1:]
-		if graph.files[name] != nil {
-			continue
-		}
 		data, ok := parts[name]
 		if ok && generated[name].stagedLeaf(data) {
 			graph.files[name], graph.used[name] = &versionEditFile{}, true
@@ -141,7 +140,8 @@ func (r *Reader) versionEditGraph(ctx context.Context, entry string, parts map[s
 				target := resolveResourcePath(name, base, value)
 				file.links = append(file.links, versionEditLink{node: node, attribute: attribute, target: target})
 				graph.used[target] = true
-				if kind == packageReferenceXML {
+				if kind == packageReferenceXML && !queued[target] {
+					queued[target] = true
 					queue = append(queue, target)
 				}
 			}
@@ -178,11 +178,11 @@ func (r *Reader) versionEditGraph(ctx context.Context, entry string, parts map[s
 // 返回: map[string]bool 受保护路径, error 读取或结构错误
 func (r *Reader) versionSharedFiles(ctx context.Context) (map[string]bool, error) {
 	shared := make(map[string]bool)
-	data, err := r.readFile("OFD.xml")
+	data, err := r.readFileViewContext(ctx, "OFD.xml")
 	if err != nil {
 		return nil, err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return nil, err
 	}
@@ -196,21 +196,25 @@ func (r *Reader) versionSharedFiles(ctx context.Context) (map[string]bool, error
 			shared[name] = true
 		}
 	}
+	var indexes []int
 	for index := range r.DocumentCount() {
-		if index == r.documentIndex {
-			continue
+		if index != r.documentIndex {
+			indexes = append(indexes, index)
 		}
-		files, err := r.documentFiles(ctx, index)
-		if err != nil {
-			return nil, err
-		}
-		maps.Copy(shared, files)
 	}
-	versions, err := r.DocumentVersions()
+	files, err := r.documentFiles(ctx, indexes...)
+	if err != nil {
+		return nil, err
+	}
+	maps.Copy(shared, files)
+	versions, err := r.DocumentVersionsContext(ctx)
 	if err != nil {
 		return nil, err
 	}
 	for _, version := range versions {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		if r.versionInfo != nil && version.ID == r.versionInfo.ID {
 			continue
 		}
@@ -253,16 +257,17 @@ func (r *Reader) versionResourceFiles(ctx context.Context, parts map[string][]by
 			files[file.Location] = true
 		}
 	}
+	var indexes []int
 	for index := range r.DocumentCount() {
-		if index == r.documentIndex {
-			continue
+		if index != r.documentIndex {
+			indexes = append(indexes, index)
 		}
-		other, err := r.documentFiles(ctx, index)
-		if err != nil {
-			return nil, err
-		}
-		maps.Copy(files, other)
 	}
+	other, err := r.documentFiles(ctx, indexes...)
+	if err != nil {
+		return nil, err
+	}
+	maps.Copy(files, other)
 	return files, nil
 }
 
@@ -273,6 +278,54 @@ func (r *Reader) versionChanges(ctx context.Context, changes map[string][]byte)
 	return r.versionOutputChanges(ctx, changes, nil, nil)
 }
 
+// versionFileUnchanged 分块比较原文件与待写入内容，读取失败时保留替换内容
+// 入参: ctx 取消上下文, name 包内路径, data 待写入内容, buffer 非空比较缓冲区
+// 返回: bool 内容相同, error 取消错误
+func (r *Reader) versionFileUnchanged(ctx context.Context, name string, data, buffer []byte) (bool, error) {
+	if err := ctx.Err(); err != nil {
+		return false, err
+	}
+	name = cleanPackagePath(name)
+	if original, ok := r.files[name]; ok {
+		if len(original) != len(data) {
+			return false, nil
+		}
+		for offset := 0; offset < len(data); {
+			if err := ctx.Err(); err != nil {
+				return false, err
+			}
+			end := offset + min(len(buffer), len(data)-offset)
+			if !bytes.Equal(original[offset:end], data[offset:end]) {
+				return false, nil
+			}
+			offset = end
+		}
+		return true, ctx.Err()
+	}
+	file, ok := r.packageFile(name)
+	if !ok || file.UncompressedSize64 != uint64(len(data)) {
+		return false, nil
+	}
+	input, err := file.Open()
+	if err != nil {
+		return false, nil
+	}
+	defer input.Close()
+	reader := imageInput{ReadCloser: input, context: ctx}
+	for len(data) != 0 {
+		block := buffer[:min(len(buffer), len(data))]
+		if _, err := io.ReadFull(&reader, block); err != nil {
+			return false, ctx.Err()
+		}
+		if !bytes.Equal(block, data[:len(block)]) {
+			return false, nil
+		}
+		data = data[len(block):]
+	}
+	n, err := reader.Read(buffer[:1])
+	return n == 0 && err == io.EOF, ctx.Err()
+}
+
 // versionOutputChanges 隔离文档或版本改动并同步删除清单，保留其他入口仍需使用的文件
 // 入参: ctx 取消上下文, changes 待写入条目, removed 待删除条目，返回前移除受保护路径, generated 本次自产页面及引用，随隔离路径更新
 // 返回: map[string][]byte 隔离后的修改集, error 读取、引用或结构错误
@@ -291,11 +344,16 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 	if parts == nil {
 		parts = make(map[string][]byte)
 	}
+	buffer := make([]byte, 32*1024)
 	for name, data := range parts {
 		if generated[name].stagedLeaf(data) {
 			continue
 		}
-		if original, err := r.readFileView(name); err == nil && bytes.Equal(original, data) {
+		same, err := r.versionFileUnchanged(ctx, name, data, buffer)
+		if err != nil {
+			return nil, err
+		}
+		if same {
 			delete(parts, name)
 		}
 	}
@@ -344,17 +402,26 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 		pending = append(pending, name)
 	}
 	for _, name := range slices.Sorted(maps.Keys(parts)) {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		if name != "OFD.xml" && shared[name] && owned[name] {
 			allocate(name)
 		}
 	}
 	parents := make(map[string][]string)
 	for _, name := range slices.Sorted(maps.Keys(graph.files)) {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		for _, link := range graph.files[name].links {
 			parents[link.target] = append(parents[link.target], name)
 		}
 	}
 	for i := 0; i < len(pending); i++ {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		for _, name := range parents[pending[i]] {
 			if shared[name] && paths[name] == "" {
 				allocate(name)
@@ -362,7 +429,7 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 		}
 	}
 	for name, file := range graph.files {
-		data, err := file.rewrite(paths)
+		data, err := file.rewrite(ctx, paths)
 		if err != nil {
 			return nil, err
 		}
@@ -373,7 +440,7 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 	for source, target := range paths {
 		data, ok := parts[source]
 		if !ok {
-			data, err = r.readFile(source)
+			data, err = r.readFileViewContext(ctx, source)
 			if err != nil {
 				return nil, err
 			}
@@ -382,18 +449,18 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 		delete(parts, source)
 	}
 	if r.versionInfo != nil {
-		if err := r.versionManifest(parts, paths, graph.used, shared, removed); err != nil {
+		if err := r.versionManifest(ctx, parts, paths, graph.used, shared, removed); err != nil {
 			return nil, err
 		}
 	} else if target := paths[entry]; target != "" {
 		data, ok := parts["OFD.xml"]
 		if !ok {
-			data, err = r.readFile("OFD.xml")
+			data, err = r.readFileViewContext(ctx, "OFD.xml")
 			if err != nil {
 				return nil, err
 			}
 		}
-		root, err := parseEditorXML(data)
+		root, err := parseEditorXMLContext(ctx, data)
 		if err != nil {
 			return nil, err
 		}
@@ -418,12 +485,21 @@ func (r *Reader) versionOutputChanges(ctx context.Context, changes map[string][]
 }
 
 // rewrite 更新已隔离文件的引用，保留其余XML及相对路径基准
-// 入参: paths 原路径到新路径的映射
+// 入参: ctx 取消上下文, paths 原路径到新路径的映射
 // 返回: []byte 更新后的XML，未改动时借用原文, error 属性编码错误
-func (file *versionEditFile) rewrite(paths map[string]string) ([]byte, error) {
+func (file *versionEditFile) rewrite(ctx context.Context, paths map[string]string) ([]byte, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	if len(paths) == 0 {
+		return file.data, nil
+	}
 	var patches []editorXMLPatch
 	headers := make(map[*editorXML][]versionEditLink)
 	for _, link := range file.links {
+		if err := ctx.Err(); err != nil {
+			return nil, err
+		}
 		if paths[link.target] == "" {
 			continue
 		}
@@ -436,7 +512,7 @@ func (file *versionEditFile) rewrite(paths map[string]string) ([]byte, error) {
 	for node, links := range headers {
 		data := file.data[node.start:node.end]
 		for _, link := range links {
-			root, err := parseEditorXML(data)
+			root, err := parseEditorXMLContext(ctx, data)
 			if err != nil {
 				return nil, err
 			}
@@ -445,7 +521,7 @@ func (file *versionEditFile) rewrite(paths map[string]string) ([]byte, error) {
 				return nil, err
 			}
 		}
-		root, err := parseEditorXML(data)
+		root, err := parseEditorXMLContext(ctx, data)
 		if err != nil {
 			return nil, err
 		}
@@ -462,15 +538,15 @@ func (file *versionEditFile) rewrite(paths map[string]string) ([]byte, error) {
 }
 
 // versionManifest 更新所选版本的入口与文件清单，保留已有文件标识及扩展字段
-// 入参: parts 输出条目, paths 隔离路径, used 当前内容依赖, shared 其他入口依赖, removed 当前版本移除的文件
+// 入参: ctx 取消上下文, parts 输出条目, paths 隔离路径, used 当前内容依赖, shared 其他入口依赖, removed 当前版本移除的文件
 // 返回: error 描述读取或XML编码错误
-func (r *Reader) versionManifest(parts map[string][]byte, paths map[string]string, used, shared, removed map[string]bool) error {
+func (r *Reader) versionManifest(ctx context.Context, parts map[string][]byte, paths map[string]string, used, shared, removed map[string]bool) error {
 	version := r.versionInfo
-	data, err := r.readFile(version.Location)
+	data, err := r.readFileViewContext(ctx, version.Location)
 	if err != nil {
 		return err
 	}
-	root, err := parseEditorXML(data)
+	root, err := parseEditorXMLContext(ctx, data)
 	if err != nil {
 		return err
 	}
@@ -479,6 +555,9 @@ func (r *Reader) versionManifest(parts map[string][]byte, paths map[string]strin
 	listed := make(map[string]bool)
 	ids := map[string]bool{version.ID: true}
 	for _, node := range list.children {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
 		if !node.matchesOFD("File") {
 			continue
 		}
@@ -500,6 +579,9 @@ func (r *Reader) versionManifest(parts map[string][]byte, paths map[string]strin
 	var added []byte
 	serial := 0
 	for _, name := range slices.Sorted(maps.Keys(used)) {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
 		if target := paths[name]; target != "" {
 			name = target
 		}
@@ -533,12 +615,12 @@ func (r *Reader) versionManifest(parts map[string][]byte, paths map[string]strin
 		name = packageAvailableName(r, parts, name)
 		entry, ok := parts["OFD.xml"]
 		if !ok {
-			entry, err = r.readFile("OFD.xml")
+			entry, err = r.readFileViewContext(ctx, "OFD.xml")
 			if err != nil {
 				return err
 			}
 		}
-		root, err := parseEditorXML(entry)
+		root, err := parseEditorXMLContext(ctx, entry)
 		if err != nil {
 			return err
 		}
diff --git a/ofdgo_writer.go b/ofdgo_writer.go
index 27bfc9c..0434fe9 100644
--- a/ofdgo_writer.go
+++ b/ofdgo_writer.go
@@ -31,7 +31,7 @@ import (
 	"time"
 )
 
-// editorProgress 保存准备阶段的进度检查点，内存预览使用nil
+// editorProgress 编辑与保存阶段的进度回调，nil表示不回报进度
 type editorProgress func(stage string, completed, total int) error
 
 // ofdCountingWriter 记录实际输出字节数
@@ -70,18 +70,22 @@ func (e *Editor) WriteTo(writer io.Writer) (int64, error) {
 
 // WritePagesTo 按指定顺序将页面另存为独立OFD，不修改当前文档或撤销记录
 // 保留描述信息并生成新文档标识，迁移相关资源、目录、注释和签章外观，移除指向未选页面的跳转
-// 快照只准备所选页面，关联引用由页面导入流程统一筛选
+// 快照只准备所选页面，关联引用由页面导入流程统一筛选，沿用当前后端及字体来源
 // 签名凭据不代表另存文件验签有效；进度与取消复用OnWriteProgress，另含snapshot、ids和commit阶段
 // 入参: writer 输出流, indexes 从0开始的页面索引，不可重复或为空
 // 返回: int64 已写入字节数, error 错误信息
 func (e *Editor) WritePagesTo(writer io.Writer, indexes []int) (int64, error) {
+	ctx := context.Background()
+	if e.output != nil {
+		ctx = e.output.ctx
+	}
 	if len(indexes) == 0 {
 		return 0, fmt.Errorf("no pages selected")
 	}
 	if err := e.validatePageIndexes(indexes); err != nil {
 		return 0, err
 	}
-	progress := editorProgress(e.OnWriteProgress)
+	progress := editorProgress(e.OnWriteProgress).withContext(ctx)
 	if err := progress.report("snapshot", 0, 0); err != nil {
 		return 0, err
 	}
@@ -104,18 +108,24 @@ func (e *Editor) WritePagesTo(writer io.Writer, indexes []int) (int64, error) {
 		snapshot.pages[i] = e.pages[index]
 		positions[i] = i
 	}
+	if err := snapshot.compressPatternPaints(); err != nil {
+		return 0, err
+	}
 	reader, err := snapshot.reader(progress)
 	if err != nil {
 		return 0, err
 	}
 	defer reader.Close()
 	selected := NewEditor()
+	selected.SetRenderBackends(e.backends)
+	selected.SetFontDirs(e.fontDirs...)
+	selected.SetFontFS(e.fontFS...)
 	selected.encryption = e.encryption
 	id := selected.Info.DocID
 	selected.Info = cloneEditorData(e.Info)
 	selected.Info.DocID = id
 	selected.OnWriteProgress = e.OnWriteProgress
-	if _, err := selected.ImportPagesWithOptions(reader, positions, 0, PageImportOptions{Outlines: true, OnProgress: e.OnWriteProgress}); err != nil {
+	if _, err := selected.ImportPagesContext(ctx, reader, positions, 0, PageImportOptions{Outlines: true, OnProgress: e.OnWriteProgress}); err != nil {
 		return 0, err
 	}
 	if e.output != nil {
@@ -153,6 +163,21 @@ func (progress editorProgress) report(stage string, completed, total int) error
 	return nil
 }
 
+// withContext 在进度回调前后检查取消，保留回调的原始错误
+// 入参: ctx 取消上下文
+// 返回: editorProgress 可取消的进度回调
+func (progress editorProgress) withContext(ctx context.Context) editorProgress {
+	return func(stage string, completed, total int) error {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+		if err := progress.report(stage, completed, total); err != nil {
+			return err
+		}
+		return ctx.Err()
+	}
+}
+
 // writePlaintext 写出未加密包，供显式明文输出与加密封装共用
 // 入参: writer 输出流
 // 返回: int64 写入字节数, error 错误信息
@@ -220,11 +245,26 @@ func (e *Editor) writePlaintext(writer io.Writer) (int64, error) {
 // 入参: progress 准备进度回调
 // 返回: *Reader 阅读器, error 错误信息
 func (e *Editor) reader(progress editorProgress) (*Reader, error) {
+	ctx := context.Background()
+	if e.output != nil {
+		ctx = e.output.ctx
+	}
+	return e.readerContext(ctx, progress)
+}
+
+// readerContext 准备可取消的内存快照，版本隔离沿用调用方上下文
+// 入参: ctx 取消上下文, progress 准备进度回调
+// 返回: *Reader 独立快照, error 准备或取消错误
+func (e *Editor) readerContext(ctx context.Context, progress editorProgress) (*Reader, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	progress = progress.withContext(ctx)
 	if err := e.validate(); err != nil {
 		return nil, err
 	}
 	if e.source != nil {
-		return e.sourceReader(progress)
+		return e.sourceReader(ctx, progress)
 	}
 	r := &Reader{files: make(map[string][]byte), encryption: e.encryption}
 	if err := e.writeParts(func(name string, data []byte, _ bool) error {
@@ -373,7 +413,7 @@ func (e *Editor) writePartsWithPages(write func(string, []byte, bool) error, pro
 		if len(fonts)+len(images)+len(spaces) != 0 {
 			x.text("DocumentRes", "DocumentRes.xml")
 		}
-		for _, name := range definitions {
+		for _, name := range slices.Sorted(maps.Keys(definitions)) {
 			x.text("DocumentRes", "/"+name)
 		}
 		x.end("CommonData")
@@ -432,7 +472,7 @@ func (e *Editor) writePartsWithPages(write func(string, []byte, bool) error, pro
 		}
 	}
 	for _, resource := range e.resources {
-		if resource.definition() != "" && slices.Contains(definitions, resource.name) || resource.colorProfileUsed(spaces) {
+		if resource.definition() != "" && definitions[resource.name] || resource.colorProfileUsed(spaces) {
 			if err := write(resource.name, resource.data, false); err != nil {
 				return err
 			}
@@ -442,8 +482,8 @@ func (e *Editor) writePartsWithPages(write func(string, []byte, bool) error, pro
 }
 
 // usedResources 沿图层、对象及资源依赖筛选新增资源和共享文件，不修改资源池
-// 返回: []editorResource 新增字体, []editorResource 新增图片, []ColorSpace 新增颜色空间, []string 独立资源文件
-func (e *Editor) usedResources() (fonts, images []editorResource, spaces []ColorSpace, sourceFiles []string) {
+// 返回: []editorResource 新增字体, []editorResource 新增图片, []ColorSpace 新增颜色空间, map[string]bool 独立资源文件集合
+func (e *Editor) usedResources() (fonts, images []editorResource, spaces []ColorSpace, sourceFiles map[string]bool) {
 	used := make(map[string]bool)
 	promoted := make(map[string]bool)
 	files := make(map[string]bool)
@@ -573,7 +613,7 @@ func (e *Editor) usedResources() (fonts, images []editorResource, spaces []Color
 			files[name] = true
 		}
 	}
-	return fonts, images, spaces, slices.Sorted(maps.Keys(files))
+	return fonts, images, spaces, files
 }
 
 // page 写出页面尺寸、图层和对象